HRNet-Semantic-Segmentation performance gap of HRNetV2+OCR on cityscape val set using default config

Hi, I'm trying to replicate the performance listed on the project page "HRNetV2-W48 + OCR val mIoU 81.6" on cityscape val set using the config file provided, i.e., "seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml". However, I can only get "Best_mIoU: 0.8033" which is ~1.3% lower than reported. Just wondering if there is any config setting I missed or extra train data was used in the reported result. Any help will be much appreciated. I noticed a previous open issue about class balance setting for the performance gap on cityscape using HRNetV2. Not sure if it is related. https://github.com/HRNet/HRNet-Semantic-Segmentation/issues/67

Jan 23 '20 04:01 verymadmatt

Hi. Would you please share your training log? We want to make sure if there's any difference.

Jan 24 '20 02:01 hsfzxjy

@hsfzxjy thanks for your quick response. the following is the training log. the training resumed from epoch 460.

2020-01-23 08:38:18,729 Namespace(cfg='experiments/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml', local_rank=3, opts=[], seed=304) 2020-01-23 08:38:18,729 Namespace(cfg='experiments/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml', local_rank=2, opts=[], seed=304) 2020-01-23 08:38:18,729 Namespace(cfg='experiments/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml', local_rank=1, opts=[], seed=304) 2020-01-23 08:38:18,729 Namespace(cfg='experiments/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml', local_rank=0, opts=[], seed=304) 2020-01-23 08:38:18,729 AUTO_RESUME: False CUDNN: BENCHMARK: True DETERMINISTIC: False ENABLED: True DATASET: DATASET: cityscapes EXTRA_TRAIN_SET: NUM_CLASSES: 19 ROOT: data/ TEST_SET: list/cityscapes/val.lst TRAIN_SET: list/cityscapes/train.lst DEBUG: DEBUG: False SAVE_BATCH_IMAGES_GT: False SAVE_BATCH_IMAGES_PRED: False SAVE_HEATMAPS_GT: False SAVE_HEATMAPS_PRED: False GPUS: (0, 1, 2, 3) LOG_DIR: log LOSS: BALANCE_WEIGHTS: [0.4, 1] CLASS_BALANCE: False OHEMKEEP: 131072 OHEMTHRES: 0.9 USE_OHEM: False MODEL: ALIGN_CORNERS: True EXTRA: FINAL_CONV_KERNEL: 1 STAGE1: BLOCK: BOTTLENECK FUSE_METHOD: SUM NUM_BLOCKS: [4] NUM_CHANNELS: [64] NUM_MODULES: 1 NUM_RANCHES: 1 STAGE2: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4] NUM_BRANCHES: 2 NUM_CHANNELS: [48, 96] NUM_MODULES: 1 STAGE3: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4, 4] NUM_BRANCHES: 3 NUM_CHANNELS: [48, 96, 192] NUM_MODULES: 4 STAGE4: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4, 4, 4] NUM_BRANCHES: 4 NUM_CHANNELS: [48, 96, 192, 384] NUM_MODULES: 3 NAME: seg_hrnet_ocr NUM_OUTPUTS: 2 OCR: DROPOUT: 0.05 KEY_CHANNELS: 256 MID_CHANNELS: 512 SCALE: 1 PRETRAINED: pretrained_models/hrnetv2_w48_imagenet_pretrained.pth OUTPUT_DIR: output PIN_MEMORY: True PRINT_FREQ: 10 RANK: 0 TEST: BASE_SIZE: 2048 BATCH_SIZE_PER_GPU: 4 FLIP_TEST: False IMAGE_SIZE: [2048, 1024] MODEL_FILE: MULTI_SCALE: False NUM_SAMPLES: 0 OUTPUT_INDEX: -1 SCALE_LIST: [1] TRAIN: BASE_SIZE: 2048 BATCH_SIZE_PER_GPU: 3 BEGIN_EPOCH: 0 DOWNSAMPLERATE: 1 END_EPOCH: 484 EXTRA_EPOCH: 0 EXTRA_LR: 0.001 FLIP: True FREEZE_EPOCHS: -1 FREEZE_LAYERS: IGNORE_LABEL: 255 IMAGE_SIZE: [1024, 512] LR: 0.01 LR_FACTOR: 0.1 LR_STEP: [90, 110] MOMENTUM: 0.9 MULTI_SCALE: True NESTEROV: False NONBACKBONE_KEYWORDS: [] NONBACKBONE_MULT: 10 NUM_SAMPLES: 0 OPTIMIZER: sgd RANDOM_BRIGHTNESS: False RANDOM_BRIGHTNESS_SHIFT_VALUE: 10 RESUME: True SCALE_FACTOR: 16 SHUFFLE: True WD: 0.0005 WORKERS: 4 2020-01-23 08:38:18,729 AUTO_RESUME: False CUDNN: BENCHMARK: True DETERMINISTIC: False ENABLED: True DATASET: DATASET: cityscapes EXTRA_TRAIN_SET: NUM_CLASSES: 19 ROOT: data/ TEST_SET: list/cityscapes/val.lst TRAIN_SET: list/cityscapes/train.lst DEBUG: DEBUG: False SAVE_BATCH_IMAGES_GT: False SAVE_BATCH_IMAGES_PRED: False SAVE_HEATMAPS_GT: False SAVE_HEATMAPS_PRED: False GPUS: (0, 1, 2, 3) LOG_DIR: log LOSS: BALANCE_WEIGHTS: [0.4, 1] CLASS_BALANCE: False OHEMKEEP: 131072 OHEMTHRES: 0.9 USE_OHEM: False MODEL: ALIGN_CORNERS: True EXTRA: FINAL_CONV_KERNEL: 1 STAGE1: BLOCK: BOTTLENECK FUSE_METHOD: SUM NUM_BLOCKS: [4] NUM_CHANNELS: [64] NUM_MODULES: 1 NUM_RANCHES: 1 STAGE2: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4] NUM_BRANCHES: 2 NUM_CHANNELS: [48, 96] NUM_MODULES: 1 STAGE3: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4, 4] NUM_BRANCHES: 3 NUM_CHANNELS: [48, 96, 192] NUM_MODULES: 4 STAGE4: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4, 4, 4] NUM_BRANCHES: 4 NUM_CHANNELS: [48, 96, 192, 384] NUM_MODULES: 3 NAME: seg_hrnet_ocr NUM_OUTPUTS: 2 OCR: DROPOUT: 0.05 KEY_CHANNELS: 256 MID_CHANNELS: 512 SCALE: 1 PRETRAINED: pretrained_models/hrnetv2_w48_imagenet_pretrained.pth OUTPUT_DIR: output PIN_MEMORY: True PRINT_FREQ: 10 RANK: 0 TEST: BASE_SIZE: 2048 BATCH_SIZE_PER_GPU: 4 FLIP_TEST: False IMAGE_SIZE: [2048, 1024] MODEL_FILE: MULTI_SCALE: False NUM_SAMPLES: 0 OUTPUT_INDEX: -1 SCALE_LIST: [1] TRAIN: BASE_SIZE: 2048 BATCH_SIZE_PER_GPU: 3 BEGIN_EPOCH: 0 DOWNSAMPLERATE: 1 END_EPOCH: 484 EXTRA_EPOCH: 0 EXTRA_LR: 0.001 FLIP: True FREEZE_EPOCHS: -1 FREEZE_LAYERS: IGNORE_LABEL: 255 IMAGE_SIZE: [1024, 512] LR: 0.01 LR_FACTOR: 0.1 LR_STEP: [90, 110] MOMENTUM: 0.9 MULTI_SCALE: True NESTEROV: False NONBACKBONE_KEYWORDS: [] NONBACKBONE_MULT: 10 NUM_SAMPLES: 0 OPTIMIZER: sgd RANDOM_BRIGHTNESS: False RANDOM_BRIGHTNESS_SHIFT_VALUE: 10 RESUME: True SCALE_FACTOR: 16 SHUFFLE: True WD: 0.0005 WORKERS: 4 2020-01-23 08:38:18,729 AUTO_RESUME: False CUDNN: BENCHMARK: True DETERMINISTIC: False ENABLED: True DATASET: DATASET: cityscapes EXTRA_TRAIN_SET: NUM_CLASSES: 19 ROOT: data/ TEST_SET: list/cityscapes/val.lst TRAIN_SET: list/cityscapes/train.lst DEBUG: DEBUG: False SAVE_BATCH_IMAGES_GT: False SAVE_BATCH_IMAGES_PRED: False SAVE_HEATMAPS_GT: False SAVE_HEATMAPS_PRED: False GPUS: (0, 1, 2, 3) LOG_DIR: log LOSS: BALANCE_WEIGHTS: [0.4, 1] CLASS_BALANCE: False OHEMKEEP: 131072 OHEMTHRES: 0.9 USE_OHEM: False MODEL: ALIGN_CORNERS: True EXTRA: FINAL_CONV_KERNEL: 1 STAGE1: BLOCK: BOTTLENECK FUSE_METHOD: SUM NUM_BLOCKS: [4] NUM_CHANNELS: [64] NUM_MODULES: 1 NUM_RANCHES: 1 STAGE2: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4] NUM_BRANCHES: 2 NUM_CHANNELS: [48, 96] NUM_MODULES: 1 STAGE3: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4, 4] NUM_BRANCHES: 3 NUM_CHANNELS: [48, 96, 192] NUM_MODULES: 4 STAGE4: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4, 4, 4] NUM_BRANCHES: 4 NUM_CHANNELS: [48, 96, 192, 384] NUM_MODULES: 3 NAME: seg_hrnet_ocr NUM_OUTPUTS: 2 OCR: DROPOUT: 0.05 KEY_CHANNELS: 256 MID_CHANNELS: 512 SCALE: 1 PRETRAINED: pretrained_models/hrnetv2_w48_imagenet_pretrained.pth OUTPUT_DIR: output PIN_MEMORY: True PRINT_FREQ: 10 RANK: 0 TEST: BASE_SIZE: 2048 BATCH_SIZE_PER_GPU: 4 FLIP_TEST: False IMAGE_SIZE: [2048, 1024] MODEL_FILE: MULTI_SCALE: False NUM_SAMPLES: 0 OUTPUT_INDEX: -1 SCALE_LIST: [1] TRAIN: BASE_SIZE: 2048 BATCH_SIZE_PER_GPU: 3 BEGIN_EPOCH: 0 DOWNSAMPLERATE: 1 END_EPOCH: 484 EXTRA_EPOCH: 0 EXTRA_LR: 0.001 FLIP: True FREEZE_EPOCHS: -1 FREEZE_LAYERS: IGNORE_LABEL: 255 IMAGE_SIZE: [1024, 512] LR: 0.01 LR_FACTOR: 0.1 LR_STEP: [90, 110] MOMENTUM: 0.9 MULTI_SCALE: True NESTEROV: False NONBACKBONE_KEYWORDS: [] NONBACKBONE_MULT: 10 NUM_SAMPLES: 0 OPTIMIZER: sgd RANDOM_BRIGHTNESS: False RANDOM_BRIGHTNESS_SHIFT_VALUE: 10 RESUME: True SCALE_FACTOR: 16 SHUFFLE: True WD: 0.0005 WORKERS: 4 2020-01-23 08:38:18,729 AUTO_RESUME: False CUDNN: BENCHMARK: True DETERMINISTIC: False ENABLED: True DATASET: DATASET: cityscapes EXTRA_TRAIN_SET: NUM_CLASSES: 19 ROOT: data/ TEST_SET: list/cityscapes/val.lst TRAIN_SET: list/cityscapes/train.lst DEBUG: DEBUG: False SAVE_BATCH_IMAGES_GT: False SAVE_BATCH_IMAGES_PRED: False SAVE_HEATMAPS_GT: False SAVE_HEATMAPS_PRED: False GPUS: (0, 1, 2, 3) LOG_DIR: log LOSS: BALANCE_WEIGHTS: [0.4, 1] CLASS_BALANCE: False OHEMKEEP: 131072 OHEMTHRES: 0.9 USE_OHEM: False MODEL: ALIGN_CORNERS: True EXTRA: FINAL_CONV_KERNEL: 1 STAGE1: BLOCK: BOTTLENECK FUSE_METHOD: SUM NUM_BLOCKS: [4] NUM_CHANNELS: [64] NUM_MODULES: 1 NUM_RANCHES: 1 STAGE2: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4] NUM_BRANCHES: 2 NUM_CHANNELS: [48, 96] NUM_MODULES: 1 STAGE3: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4, 4] NUM_BRANCHES: 3 NUM_CHANNELS: [48, 96, 192] NUM_MODULES: 4 STAGE4: BLOCK: BASIC FUSE_METHOD: SUM NUM_BLOCKS: [4, 4, 4, 4] NUM_BRANCHES: 4 NUM_CHANNELS: [48, 96, 192, 384] NUM_MODULES: 3 NAME: seg_hrnet_ocr NUM_OUTPUTS: 2 OCR: DROPOUT: 0.05 KEY_CHANNELS: 256 MID_CHANNELS: 512 SCALE: 1 PRETRAINED: pretrained_models/hrnetv2_w48_imagenet_pretrained.pth OUTPUT_DIR: output PIN_MEMORY: True PRINT_FREQ: 10 RANK: 0 TEST: BASE_SIZE: 2048 BATCH_SIZE_PER_GPU: 4 FLIP_TEST: False IMAGE_SIZE: [2048, 1024] MODEL_FILE: MULTI_SCALE: False NUM_SAMPLES: 0 OUTPUT_INDEX: -1 SCALE_LIST: [1] TRAIN: BASE_SIZE: 2048 BATCH_SIZE_PER_GPU: 3 BEGIN_EPOCH: 0 DOWNSAMPLERATE: 1 END_EPOCH: 484 EXTRA_EPOCH: 0 EXTRA_LR: 0.001 FLIP: True FREEZE_EPOCHS: -1 FREEZE_LAYERS: IGNORE_LABEL: 255 IMAGE_SIZE: [1024, 512] LR: 0.01 LR_FACTOR: 0.1 LR_STEP: [90, 110] MOMENTUM: 0.9 MULTI_SCALE: True NESTEROV: False NONBACKBONE_KEYWORDS: [] NONBACKBONE_MULT: 10 NUM_SAMPLES: 0 OPTIMIZER: sgd RANDOM_BRIGHTNESS: False RANDOM_BRIGHTNESS_SHIFT_VALUE: 10 RESUME: True SCALE_FACTOR: 16 SHUFFLE: True WD: 0.0005 WORKERS: 4 2020-01-23 08:38:20,239 => init weights from normal distribution 2020-01-23 08:38:20,244 => init weights from normal distribution 2020-01-23 08:38:21,243 => init weights from normal distribution 2020-01-23 08:38:21,244 => init weights from normal distribution 2020-01-23 08:38:22,839 => loading pretrained model pretrained_models/hrnetv2_w48_imagenet_pretrained.pth 2020-01-23 08:38:22,839 => loading pretrained model pretrained_models/hrnetv2_w48_imagenet_pretrained.pth 2020-01-23 08:38:22,840 => loading pretrained model pretrained_models/hrnetv2_w48_imagenet_pretrained.pth 2020-01-23 08:38:22,840 => loading pretrained model pretrained_models/hrnetv2_w48_imagenet_pretrained.pth 2020-01-23 08:38:46,009 => loaded checkpoint (epoch 460) 2020-01-23 08:38:46,013 => loaded checkpoint (epoch 460) 2020-01-23 08:38:46,019 => loaded checkpoint (epoch 460) 2020-01-23 08:38:46,022 => loaded checkpoint (epoch 460) 2020-01-23 08:38:58,242 Epoch: [460/484] Iter:[0/247], Time: 12.21, lr: [0.0006696213499130277], Loss: 0.118911 2020-01-23 08:39:07,370 Epoch: [460/484] Iter:[10/247], Time: 1.94, lr: [0.0006686046325075444], Loss: 0.119875 2020-01-23 08:39:16,070 Epoch: [460/484] Iter:[20/247], Time: 1.43, lr: [0.0006675877432866135], Loss: 0.128737 2020-01-23 08:39:24,247 Epoch: [460/484] Iter:[30/247], Time: 1.23, lr: [0.0006665706819303064], Loss: 0.135692 2020-01-23 08:39:32,488 Epoch: [460/484] Iter:[40/247], Time: 1.13, lr: [0.0006655534481175586], Loss: 0.129678 2020-01-23 08:39:40,553 Epoch: [460/484] Iter:[50/247], Time: 1.07, lr: [0.0006645360415261557], Loss: 0.127583 2020-01-23 08:39:48,696 Epoch: [460/484] Iter:[60/247], Time: 1.03, lr: [0.0006635184618327374], Loss: 0.129885 2020-01-23 08:39:56,943 Epoch: [460/484] Iter:[70/247], Time: 1.00, lr: [0.0006625007087127803], Loss: 0.127894 2020-01-23 08:40:05,347 Epoch: [460/484] Iter:[80/247], Time: 0.98, lr: [0.0006614827818406031], Loss: 0.127111 2020-01-23 08:40:13,642 Epoch: [460/484] Iter:[90/247], Time: 0.96, lr: [0.0006604646808893498], Loss: 0.126133 2020-01-23 08:40:21,899 Epoch: [460/484] Iter:[100/247], Time: 0.95, lr: [0.0006594464055309934], Loss: 0.126229 2020-01-23 08:40:30,274 Epoch: [460/484] Iter:[110/247], Time: 0.94, lr: [0.0006584279554363199], Loss: 0.125137 2020-01-23 08:40:38,471 Epoch: [460/484] Iter:[120/247], Time: 0.93, lr: [0.0006574093302749319], Loss: 0.130603 2020-01-23 08:40:46,558 Epoch: [460/484] Iter:[130/247], Time: 0.92, lr: [0.0006563905297152323], Loss: 0.130406 2020-01-23 08:40:54,672 Epoch: [460/484] Iter:[140/247], Time: 0.91, lr: [0.0006553715534244278], Loss: 0.131272 2020-01-23 08:41:02,779 Epoch: [460/484] Iter:[150/247], Time: 0.91, lr: [0.0006543524010685127], Loss: 0.129639 2020-01-23 08:41:10,782 Epoch: [460/484] Iter:[160/247], Time: 0.90, lr: [0.0006533330723122722], Loss: 0.131556 2020-01-23 08:41:18,960 Epoch: [460/484] Iter:[170/247], Time: 0.89, lr: [0.0006523135668192672], Loss: 0.131600 2020-01-23 08:41:27,032 Epoch: [460/484] Iter:[180/247], Time: 0.89, lr: [0.0006512938842518311], Loss: 0.131241 2020-01-23 08:41:35,148 Epoch: [460/484] Iter:[190/247], Time: 0.89, lr: [0.000650274024271068], Loss: 0.131862 2020-01-23 08:41:43,255 Epoch: [460/484] Iter:[200/247], Time: 0.88, lr: [0.0006492539865368359], Loss: 0.130830 2020-01-23 08:41:51,305 Epoch: [460/484] Iter:[210/247], Time: 0.88, lr: [0.0006482337707077512], Loss: 0.130380 2020-01-23 08:41:59,540 Epoch: [460/484] Iter:[220/247], Time: 0.88, lr: [0.0006472133764411711], Loss: 0.129506 2020-01-23 08:42:07,761 Epoch: [460/484] Iter:[230/247], Time: 0.87, lr: [0.0006461928033931973], Loss: 0.130382 2020-01-23 08:42:15,844 Epoch: [460/484] Iter:[240/247], Time: 0.87, lr: [0.0006451720512186586], Loss: 0.130188 2020-01-23 08:45:45,871 0 [0.98437091 0.87080627 0.93431437 0.56149059 0.63651587 0.69703517 0.7438209 0.81912795 0.92951809 0.63110144 0.95263837 0.83888669 0.64346649 0.95265834 0.70352713 0.86580247 0.70959489 0.64546466 0.79148257] 0.7848222718276417 2020-01-23 08:45:45,872 1 [0.98455361 0.87262859 0.93521555 0.55916849 0.64176191 0.70092739 0.74658187 0.82199962 0.93013398 0.63550485 0.9528006 0.84105856 0.65051016 0.95634688 0.76277092 0.88573624 0.79219625 0.65678277 0.79405576] 0.7958281055046451 2020-01-23 08:45:45,872 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 08:45:48,771 Loss: 0.164, MeanIU: 0.7958, Best_mIoU: 0.7984 2020-01-23 08:45:48,771 [0.98455361 0.87262859 0.93521555 0.55916849 0.64176191 0.70092739 0.74658187 0.82199962 0.93013398 0.63550485 0.9528006 0.84105856 0.65051016 0.95634688 0.76277092 0.88573624 0.79219625 0.65678277 0.79405576] 2020-01-23 08:45:51,088 Epoch: [461/484] Iter:[0/247], Time: 2.31, lr: [0.0006444574179307085], Loss: 0.075820 2020-01-23 08:45:59,151 Epoch: [461/484] Iter:[10/247], Time: 0.94, lr: [0.0006434363604452395], Loss: 0.116503 2020-01-23 08:46:07,345 Epoch: [461/484] Iter:[20/247], Time: 0.88, lr: [0.0006424151228948677], Loss: 0.119192 2020-01-23 08:46:15,497 Epoch: [461/484] Iter:[30/247], Time: 0.86, lr: [0.000641393704929673], Loss: 0.124967 2020-01-23 08:46:23,526 Epoch: [461/484] Iter:[40/247], Time: 0.85, lr: [0.0006403721061984345], Loss: 0.125565 2020-01-23 08:46:31,797 Epoch: [461/484] Iter:[50/247], Time: 0.84, lr: [0.0006393503263486281], Loss: 0.122569 2020-01-23 08:46:40,026 Epoch: [461/484] Iter:[60/247], Time: 0.84, lr: [0.0006383283650264092], Loss: 0.121108 2020-01-23 08:46:48,309 Epoch: [461/484] Iter:[70/247], Time: 0.84, lr: [0.0006373062218766165], Loss: 0.122991 2020-01-23 08:46:56,630 Epoch: [461/484] Iter:[80/247], Time: 0.84, lr: [0.0006362838965427542], Loss: 0.126043 2020-01-23 08:47:04,675 Epoch: [461/484] Iter:[90/247], Time: 0.83, lr: [0.0006352613886669949], Loss: 0.128042 2020-01-23 08:47:12,794 Epoch: [461/484] Iter:[100/247], Time: 0.83, lr: [0.0006342386978901619], Loss: 0.127832 2020-01-23 08:47:21,070 Epoch: [461/484] Iter:[110/247], Time: 0.83, lr: [0.0006332158238517319], Loss: 0.127493 2020-01-23 08:47:29,122 Epoch: [461/484] Iter:[120/247], Time: 0.83, lr: [0.0006321927661898176], Loss: 0.126949 2020-01-23 08:47:37,172 Epoch: [461/484] Iter:[130/247], Time: 0.83, lr: [0.0006311695245411698], Loss: 0.128273 2020-01-23 08:47:45,569 Epoch: [461/484] Iter:[140/247], Time: 0.83, lr: [0.0006301460985411603], Loss: 0.128170 2020-01-23 08:47:53,856 Epoch: [461/484] Iter:[150/247], Time: 0.83, lr: [0.0006291224878237833], Loss: 0.128914 2020-01-23 08:48:01,924 Epoch: [461/484] Iter:[160/247], Time: 0.83, lr: [0.0006280986920216388], Loss: 0.129907 2020-01-23 08:48:10,179 Epoch: [461/484] Iter:[170/247], Time: 0.83, lr: [0.0006270747107659339], Loss: 0.128707 2020-01-23 08:48:18,273 Epoch: [461/484] Iter:[180/247], Time: 0.83, lr: [0.0006260505436864653], Loss: 0.127596 2020-01-23 08:48:26,369 Epoch: [461/484] Iter:[190/247], Time: 0.83, lr: [0.0006250261904116214], Loss: 0.129863 2020-01-23 08:48:34,534 Epoch: [461/484] Iter:[200/247], Time: 0.82, lr: [0.0006240016505683654], Loss: 0.129164 2020-01-23 08:48:42,794 Epoch: [461/484] Iter:[210/247], Time: 0.82, lr: [0.000622976923782231], Loss: 0.129494 2020-01-23 08:48:50,994 Epoch: [461/484] Iter:[220/247], Time: 0.82, lr: [0.0006219520096773186], Loss: 0.128742 2020-01-23 08:48:59,207 Epoch: [461/484] Iter:[230/247], Time: 0.82, lr: [0.0006209269078762776], Loss: 0.129028 2020-01-23 08:49:07,365 Epoch: [461/484] Iter:[240/247], Time: 0.82, lr: [0.0006199016180003085], Loss: 0.128983 2020-01-23 08:52:44,292 0 [0.98403233 0.86677607 0.93275096 0.57285824 0.6380202 0.69608891 0.7419803 0.81982863 0.92918788 0.63017172 0.95188877 0.84104149 0.64715434 0.9493354 0.66410468 0.85237538 0.69260188 0.65644771 0.78923281] 0.7818883008411794 2020-01-23 08:52:44,292 1 [0.98429423 0.86882542 0.93320899 0.57588066 0.64276964 0.70052156 0.74241117 0.82194612 0.92984191 0.63438439 0.9528121 0.84287945 0.65216577 0.95139094 0.69170568 0.87519977 0.7650452 0.66595666 0.79139403] 0.7906649312589081 2020-01-23 08:52:44,293 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 08:52:47,104 Loss: 0.168, MeanIU: 0.7907, Best_mIoU: 0.7984 2020-01-23 08:52:47,105 [0.98429423 0.86882542 0.93320899 0.57588066 0.64276964 0.70052156 0.74241117 0.82194612 0.92984191 0.63438439 0.9528121 0.84287945 0.65216577 0.95139094 0.69170568 0.87519977 0.7650452 0.66595666 0.79139403] 2020-01-23 08:52:48,957 Epoch: [462/484] Iter:[0/247], Time: 1.85, lr: [0.0006191838029789736], Loss: 0.127456 2020-01-23 08:52:57,178 Epoch: [462/484] Iter:[10/247], Time: 0.92, lr: [0.0006181581925021317], Loss: 0.124150 2020-01-23 08:53:05,344 Epoch: [462/484] Iter:[20/247], Time: 0.87, lr: [0.000617132392920267], Loss: 0.125770 2020-01-23 08:53:13,614 Epoch: [462/484] Iter:[30/247], Time: 0.85, lr: [0.0006161064038491285], Loss: 0.125951 2020-01-23 08:53:21,840 Epoch: [462/484] Iter:[40/247], Time: 0.85, lr: [0.0006150802249029663], Loss: 0.125334 2020-01-23 08:53:30,035 Epoch: [462/484] Iter:[50/247], Time: 0.84, lr: [0.000614053855694534], Loss: 0.126813 2020-01-23 08:53:38,357 Epoch: [462/484] Iter:[60/247], Time: 0.84, lr: [0.0006130272958350709], Loss: 0.126591 2020-01-23 08:53:46,455 Epoch: [462/484] Iter:[70/247], Time: 0.84, lr: [0.0006120005449342964], Loss: 0.124839 2020-01-23 08:53:54,596 Epoch: [462/484] Iter:[80/247], Time: 0.83, lr: [0.0006109736026004065], Loss: 0.126375 2020-01-23 08:54:02,803 Epoch: [462/484] Iter:[90/247], Time: 0.83, lr: [0.0006099464684400544], Loss: 0.126471 2020-01-23 08:54:10,841 Epoch: [462/484] Iter:[100/247], Time: 0.83, lr: [0.0006089191420583531], Loss: 0.127334 2020-01-23 08:54:18,991 Epoch: [462/484] Iter:[110/247], Time: 0.83, lr: [0.0006078916230588552], Loss: 0.128796 2020-01-23 08:54:27,418 Epoch: [462/484] Iter:[120/247], Time: 0.83, lr: [0.0006068639110435551], Loss: 0.129228 2020-01-23 08:54:35,602 Epoch: [462/484] Iter:[130/247], Time: 0.83, lr: [0.0006058360056128684], Loss: 0.128529 2020-01-23 08:54:43,684 Epoch: [462/484] Iter:[140/247], Time: 0.83, lr: [0.000604807906365634], Loss: 0.129156 2020-01-23 08:54:52,037 Epoch: [462/484] Iter:[150/247], Time: 0.83, lr: [0.0006037796128990944], Loss: 0.129477 2020-01-23 08:55:00,322 Epoch: [462/484] Iter:[160/247], Time: 0.83, lr: [0.0006027511248088959], Loss: 0.128126 2020-01-23 08:55:08,570 Epoch: [462/484] Iter:[170/247], Time: 0.83, lr: [0.0006017224416890697], Loss: 0.127795 2020-01-23 08:55:16,846 Epoch: [462/484] Iter:[180/247], Time: 0.83, lr: [0.0006006935631320328], Loss: 0.128891 2020-01-23 08:55:25,041 Epoch: [462/484] Iter:[190/247], Time: 0.83, lr: [0.0005996644887285671], Loss: 0.128605 2020-01-23 08:55:33,194 Epoch: [462/484] Iter:[200/247], Time: 0.83, lr: [0.0005986352180678213], Loss: 0.127676 2020-01-23 08:55:41,605 Epoch: [462/484] Iter:[210/247], Time: 0.83, lr: [0.0005976057507372914], Loss: 0.128769 2020-01-23 08:55:49,835 Epoch: [462/484] Iter:[220/247], Time: 0.83, lr: [0.0005965760863228149], Loss: 0.129427 2020-01-23 08:55:58,491 Epoch: [462/484] Iter:[230/247], Time: 0.83, lr: [0.0005955462244085654], Loss: 0.129148 2020-01-23 08:56:07,723 Epoch: [462/484] Iter:[240/247], Time: 0.83, lr: [0.0005945161645770325], Loss: 0.130601 2020-01-23 09:00:06,623 0 [0.9840897 0.86739981 0.93417353 0.58468911 0.63395352 0.70207379 0.74306553 0.82401682 0.93024343 0.63247745 0.95254216 0.84086631 0.63837956 0.95037691 0.65801323 0.8286728 0.56503247 0.65961128 0.79365219] 0.7749120849113121 2020-01-23 09:00:06,624 1 [0.98432422 0.86948458 0.93441968 0.58243555 0.63745061 0.70698372 0.74554226 0.82588719 0.93075165 0.63671922 0.95261371 0.84376925 0.64656234 0.951723 0.67282398 0.82875147 0.54193775 0.66781598 0.79688137] 0.7766777658097438 2020-01-23 09:00:06,624 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 09:00:09,535 Loss: 0.167, MeanIU: 0.7767, Best_mIoU: 0.7984 2020-01-23 09:00:09,535 [0.98432422 0.86948458 0.93441968 0.58243555 0.63745061 0.70698372 0.74554226 0.82588719 0.93075165 0.63671922 0.95261371 0.84376925 0.64656234 0.951723 0.67282398 0.82875147 0.54193775 0.66781598 0.79688137] 2020-01-23 09:00:11,491 Epoch: [463/484] Iter:[0/247], Time: 1.94, lr: [0.0005937950047097665], Loss: 0.128534 2020-01-23 09:00:19,737 Epoch: [463/484] Iter:[10/247], Time: 0.93, lr: [0.0005927646074558912], Loss: 0.111006 2020-01-23 09:00:28,046 Epoch: [463/484] Iter:[20/247], Time: 0.88, lr: [0.0005917340111490039], Loss: 0.105998 2020-01-23 09:00:36,290 Epoch: [463/484] Iter:[30/247], Time: 0.86, lr: [0.0005907032153652975], Loss: 0.114645 2020-01-23 09:00:44,363 Epoch: [463/484] Iter:[40/247], Time: 0.85, lr: [0.000589672219679244], Loss: 0.117490 2020-01-23 09:00:52,443 Epoch: [463/484] Iter:[50/247], Time: 0.84, lr: [0.0005886410236635743], Loss: 0.124547 2020-01-23 09:01:00,572 Epoch: [463/484] Iter:[60/247], Time: 0.84, lr: [0.0005876096268892771], Loss: 0.127271 2020-01-23 09:01:08,627 Epoch: [463/484] Iter:[70/247], Time: 0.83, lr: [0.0005865780289255803], Loss: 0.125105 2020-01-23 09:01:16,827 Epoch: [463/484] Iter:[80/247], Time: 0.83, lr: [0.0005855462293399432], Loss: 0.126093 2020-01-23 09:01:25,035 Epoch: [463/484] Iter:[90/247], Time: 0.83, lr: [0.0005845142276980508], Loss: 0.127423 2020-01-23 09:01:33,259 Epoch: [463/484] Iter:[100/247], Time: 0.83, lr: [0.000583482023563793], Loss: 0.127719 2020-01-23 09:01:41,438 Epoch: [463/484] Iter:[110/247], Time: 0.83, lr: [0.0005824496164992644], Loss: 0.127757 2020-01-23 09:01:49,638 Epoch: [463/484] Iter:[120/247], Time: 0.83, lr: [0.0005814170060647429], Loss: 0.128981 2020-01-23 09:01:57,878 Epoch: [463/484] Iter:[130/247], Time: 0.83, lr: [0.0005803841918186891], Loss: 0.128008 2020-01-23 09:02:05,969 Epoch: [463/484] Iter:[140/247], Time: 0.83, lr: [0.0005793511733177245], Loss: 0.128694 2020-01-23 09:02:14,111 Epoch: [463/484] Iter:[150/247], Time: 0.82, lr: [0.0005783179501166311], Loss: 0.128120 2020-01-23 09:02:22,300 Epoch: [463/484] Iter:[160/247], Time: 0.82, lr: [0.0005772845217683287], Loss: 0.127989 2020-01-23 09:02:30,538 Epoch: [463/484] Iter:[170/247], Time: 0.82, lr: [0.0005762508878238752], Loss: 0.127994 2020-01-23 09:02:38,784 Epoch: [463/484] Iter:[180/247], Time: 0.82, lr: [0.0005752170478324435], Loss: 0.127578 2020-01-23 09:02:46,861 Epoch: [463/484] Iter:[190/247], Time: 0.82, lr: [0.0005741830013413207], Loss: 0.126680 2020-01-23 09:02:55,486 Epoch: [463/484] Iter:[200/247], Time: 0.83, lr: [0.0005731487478958862], Loss: 0.127300 2020-01-23 09:03:04,407 Epoch: [463/484] Iter:[210/247], Time: 0.83, lr: [0.0005721142870396096], Loss: 0.128803 2020-01-23 09:03:13,419 Epoch: [463/484] Iter:[220/247], Time: 0.83, lr: [0.0005710796183140307], Loss: 0.128047 2020-01-23 09:03:22,231 Epoch: [463/484] Iter:[230/247], Time: 0.83, lr: [0.0005700447412587501], Loss: 0.127893 2020-01-23 09:03:31,153 Epoch: [463/484] Iter:[240/247], Time: 0.84, lr: [0.0005690096554114224], Loss: 0.127927 2020-01-23 09:07:16,434 0 [0.98402281 0.86732705 0.93242174 0.54977841 0.62725479 0.69929614 0.74057547 0.82111322 0.93053242 0.64419378 0.95366707 0.84080874 0.64104054 0.95160795 0.71439027 0.86444502 0.71138762 0.6662869 0.79010774] 0.7858030359719429 2020-01-23 09:07:16,435 1 [0.98437887 0.86957622 0.93313335 0.55901336 0.63133668 0.70323038 0.74229717 0.82347385 0.93128121 0.64864622 0.95396813 0.84316724 0.64509483 0.95479021 0.7747153 0.89660925 0.80678609 0.67543924 0.79252834] 0.7983929438289701 2020-01-23 09:07:16,435 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 09:07:20,687 Loss: 0.167, MeanIU: 0.7984, Best_mIoU: 0.7984 2020-01-23 09:07:20,687 [0.98437887 0.86957622 0.93313335 0.55901336 0.63133668 0.70323038 0.74229717 0.82347385 0.93128121 0.64864622 0.95396813 0.84316724 0.64509483 0.95479021 0.7747153 0.89660925 0.80678609 0.67543924 0.79252834] 2020-01-23 09:07:22,570 Epoch: [464/484] Iter:[0/247], Time: 1.88, lr: [0.0005682849708384633], Loss: 0.140334 2020-01-23 09:07:30,645 Epoch: [464/484] Iter:[10/247], Time: 0.90, lr: [0.0005672495289780013], Loss: 0.112829 2020-01-23 09:07:38,655 Epoch: [464/484] Iter:[20/247], Time: 0.86, lr: [0.0005662138770673093], Loss: 0.124635 2020-01-23 09:07:46,815 Epoch: [464/484] Iter:[30/247], Time: 0.84, lr: [0.000565178014636717], Loss: 0.122768 2020-01-23 09:07:54,906 Epoch: [464/484] Iter:[40/247], Time: 0.83, lr: [0.00056414194121454], Loss: 0.123501 2020-01-23 09:08:02,936 Epoch: [464/484] Iter:[50/247], Time: 0.83, lr: [0.0005631056563270772], Loss: 0.125901 2020-01-23 09:08:11,022 Epoch: [464/484] Iter:[60/247], Time: 0.83, lr: [0.000562069159498588], Loss: 0.124226 2020-01-23 09:08:19,149 Epoch: [464/484] Iter:[70/247], Time: 0.82, lr: [0.0005610324502512894], Loss: 0.124301 2020-01-23 09:08:27,445 Epoch: [464/484] Iter:[80/247], Time: 0.82, lr: [0.0005599955281053328], Loss: 0.120878 2020-01-23 09:08:35,641 Epoch: [464/484] Iter:[90/247], Time: 0.82, lr: [0.0005589583925788012], Loss: 0.118658 2020-01-23 09:08:43,727 Epoch: [464/484] Iter:[100/247], Time: 0.82, lr: [0.0005579210431876868], Loss: 0.118407 2020-01-23 09:08:51,771 Epoch: [464/484] Iter:[110/247], Time: 0.82, lr: [0.0005568834794458814], Loss: 0.119367 2020-01-23 09:08:59,929 Epoch: [464/484] Iter:[120/247], Time: 0.82, lr: [0.0005558457008651673], Loss: 0.118923 2020-01-23 09:09:08,136 Epoch: [464/484] Iter:[130/247], Time: 0.82, lr: [0.0005548077069551938], Loss: 0.118975 2020-01-23 09:09:16,165 Epoch: [464/484] Iter:[140/247], Time: 0.82, lr: [0.0005537694972234748], Loss: 0.119314 2020-01-23 09:09:24,224 Epoch: [464/484] Iter:[150/247], Time: 0.82, lr: [0.000552731071175364], Loss: 0.120334 2020-01-23 09:09:32,198 Epoch: [464/484] Iter:[160/247], Time: 0.82, lr: [0.0005516924283140512], Loss: 0.125043 2020-01-23 09:09:40,379 Epoch: [464/484] Iter:[170/247], Time: 0.82, lr: [0.0005506535681405385], Loss: 0.125799 2020-01-23 09:09:48,521 Epoch: [464/484] Iter:[180/247], Time: 0.82, lr: [0.0005496144901536361], Loss: 0.128367 2020-01-23 09:09:56,747 Epoch: [464/484] Iter:[190/247], Time: 0.82, lr: [0.0005485751938499374], Loss: 0.127717 2020-01-23 09:10:04,945 Epoch: [464/484] Iter:[200/247], Time: 0.82, lr: [0.0005475356787238154], Loss: 0.128366 2020-01-23 09:10:13,053 Epoch: [464/484] Iter:[210/247], Time: 0.82, lr: [0.0005464959442673973], Loss: 0.127815 2020-01-23 09:10:21,246 Epoch: [464/484] Iter:[220/247], Time: 0.82, lr: [0.0005454559899705605], Loss: 0.126646 2020-01-23 09:10:29,445 Epoch: [464/484] Iter:[230/247], Time: 0.82, lr: [0.0005444158153209075], Loss: 0.126749 2020-01-23 09:10:37,556 Epoch: [464/484] Iter:[240/247], Time: 0.82, lr: [0.0005433754198037612], Loss: 0.127773 2020-01-23 09:14:13,899 0 [0.98399046 0.86773851 0.93315325 0.60056559 0.6363511 0.70055369 0.73813424 0.81897529 0.92984023 0.64331609 0.95247502 0.84191888 0.64395434 0.95419988 0.74980507 0.8684766 0.733729 0.66361353 0.79496267] 0.7924080749396682 2020-01-23 09:14:13,900 1 [0.98412237 0.86899429 0.93281983 0.5759732 0.63895571 0.70488978 0.73952183 0.81958489 0.93063633 0.64819003 0.95322453 0.8441142 0.64981107 0.95804859 0.82619482 0.8946129 0.82476427 0.67021678 0.79839178] 0.8033193267783957 2020-01-23 09:14:13,901 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 09:14:18,136 Loss: 0.162, MeanIU: 0.8033, Best_mIoU: 0.8033 2020-01-23 09:14:18,136 [0.98412237 0.86899429 0.93281983 0.5759732 0.63895571 0.70488978 0.73952183 0.81958489 0.93063633 0.64819003 0.95322453 0.8441142 0.64981107 0.95804859 0.82619482 0.8946129 0.82476427 0.67021678 0.79839178] 2020-01-23 09:14:19,945 Epoch: [465/484] Iter:[0/247], Time: 1.80, lr: [0.0005426470112488497], Loss: 0.119499 2020-01-23 09:14:28,105 Epoch: [465/484] Iter:[10/247], Time: 0.91, lr: [0.0005416062390692534], Loss: 0.121959 2020-01-23 09:14:36,346 Epoch: [465/484] Iter:[20/247], Time: 0.87, lr: [0.0005405652446209973], Loss: 0.122974 2020-01-23 09:14:44,600 Epoch: [465/484] Iter:[30/247], Time: 0.85, lr: [0.0005395240273808128], Loss: 0.123857 2020-01-23 09:14:52,682 Epoch: [465/484] Iter:[40/247], Time: 0.84, lr: [0.0005384825868230803], Loss: 0.132320 2020-01-23 09:15:00,766 Epoch: [465/484] Iter:[50/247], Time: 0.84, lr: [0.000537440922419802], Loss: 0.129501 2020-01-23 09:15:08,827 Epoch: [465/484] Iter:[60/247], Time: 0.83, lr: [0.0005363990336405973], Loss: 0.131007 2020-01-23 09:15:16,814 Epoch: [465/484] Iter:[70/247], Time: 0.83, lr: [0.0005353569199526755], Loss: 0.128086 2020-01-23 09:15:25,095 Epoch: [465/484] Iter:[80/247], Time: 0.83, lr: [0.0005343145808208316], Loss: 0.129547 2020-01-23 09:15:33,060 Epoch: [465/484] Iter:[90/247], Time: 0.82, lr: [0.0005332720157074176], Loss: 0.127596 2020-01-23 09:15:41,295 Epoch: [465/484] Iter:[100/247], Time: 0.82, lr: [0.0005322292240723384], Loss: 0.130999 2020-01-23 09:15:49,231 Epoch: [465/484] Iter:[110/247], Time: 0.82, lr: [0.0005311862053730249], Loss: 0.130745 2020-01-23 09:15:57,399 Epoch: [465/484] Iter:[120/247], Time: 0.82, lr: [0.0005301429590644216], Loss: 0.130638 2020-01-23 09:16:05,404 Epoch: [465/484] Iter:[130/247], Time: 0.82, lr: [0.0005290994845989738], Loss: 0.130766 2020-01-23 09:16:13,711 Epoch: [465/484] Iter:[140/247], Time: 0.82, lr: [0.0005280557814266012], Loss: 0.131708 2020-01-23 09:16:21,841 Epoch: [465/484] Iter:[150/247], Time: 0.82, lr: [0.0005270118489946914], Loss: 0.131902 2020-01-23 09:16:30,076 Epoch: [465/484] Iter:[160/247], Time: 0.82, lr: [0.0005259676867480717], Loss: 0.132308 2020-01-23 09:16:38,199 Epoch: [465/484] Iter:[170/247], Time: 0.82, lr: [0.000524923294129003], Loss: 0.131581 2020-01-23 09:16:46,410 Epoch: [465/484] Iter:[180/247], Time: 0.82, lr: [0.0005238786705771505], Loss: 0.131287 2020-01-23 09:16:54,370 Epoch: [465/484] Iter:[190/247], Time: 0.82, lr: [0.0005228338155295776], Loss: 0.130705 2020-01-23 09:17:02,726 Epoch: [465/484] Iter:[200/247], Time: 0.82, lr: [0.000521788728420716], Loss: 0.130647 2020-01-23 09:17:10,807 Epoch: [465/484] Iter:[210/247], Time: 0.82, lr: [0.0005207434086823599], Loss: 0.130546 2020-01-23 09:17:18,939 Epoch: [465/484] Iter:[220/247], Time: 0.82, lr: [0.0005196978557436356], Loss: 0.129339 2020-01-23 09:17:27,183 Epoch: [465/484] Iter:[230/247], Time: 0.82, lr: [0.0005186520690309946], Loss: 0.129609 2020-01-23 09:17:35,273 Epoch: [465/484] Iter:[240/247], Time: 0.82, lr: [0.0005176060479681835], Loss: 0.129010 2020-01-23 09:21:15,928 0 [0.98400896 0.86588743 0.93322819 0.5863259 0.63489848 0.69568069 0.73925896 0.82185804 0.93054229 0.62927777 0.95392777 0.84044663 0.64045535 0.949831 0.65139938 0.85477072 0.66746655 0.65996037 0.79248751] 0.7806164206063521 2020-01-23 09:21:15,929 1 [0.98417145 0.86733762 0.93364575 0.57548772 0.64029347 0.69941482 0.74183735 0.82471449 0.93124318 0.63495368 0.95471954 0.8423293 0.64127981 0.95193913 0.68417083 0.87891298 0.7500817 0.66462009 0.79464445] 0.7892524930901829 2020-01-23 09:21:15,929 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 09:21:18,854 Loss: 0.164, MeanIU: 0.7893, Best_mIoU: 0.8033 2020-01-23 09:21:18,855 [0.98417145 0.86733762 0.93364575 0.57548772 0.64029347 0.69941482 0.74183735 0.82471449 0.93124318 0.63495368 0.95471954 0.8423293 0.64127981 0.95193913 0.68417083 0.87891298 0.7500817 0.66462009 0.79464445] 2020-01-23 09:21:20,728 Epoch: [466/484] Iter:[0/247], Time: 1.87, lr: [0.0005168736934759387], Loss: 0.078918 2020-01-23 09:21:28,857 Epoch: [466/484] Iter:[10/247], Time: 0.91, lr: [0.0005158272726876455], Loss: 0.130154 2020-01-23 09:21:36,911 Epoch: [466/484] Iter:[20/247], Time: 0.86, lr: [0.0005147806159796768], Loss: 0.137728 2020-01-23 09:21:45,193 Epoch: [466/484] Iter:[30/247], Time: 0.85, lr: [0.0005137337227656337], Loss: 0.134108 2020-01-23 09:21:53,278 Epoch: [466/484] Iter:[40/247], Time: 0.84, lr: [0.0005126865924563228], Loss: 0.138695 2020-01-23 09:22:01,391 Epoch: [466/484] Iter:[50/247], Time: 0.83, lr: [0.0005116392244597475], Loss: 0.138556 2020-01-23 09:22:09,481 Epoch: [466/484] Iter:[60/247], Time: 0.83, lr: [0.0005105916181810776], Loss: 0.138651 2020-01-23 09:22:17,771 Epoch: [466/484] Iter:[70/247], Time: 0.83, lr: [0.0005095437730226401], Loss: 0.137149 2020-01-23 09:22:25,825 Epoch: [466/484] Iter:[80/247], Time: 0.83, lr: [0.0005084956883838877], Loss: 0.137584 2020-01-23 09:22:33,918 Epoch: [466/484] Iter:[90/247], Time: 0.82, lr: [0.00050744736366139], Loss: 0.135971 2020-01-23 09:22:41,981 Epoch: [466/484] Iter:[100/247], Time: 0.82, lr: [0.0005063987982488014], Loss: 0.135384 2020-01-23 09:22:50,059 Epoch: [466/484] Iter:[110/247], Time: 0.82, lr: [0.0005053499915368516], Loss: 0.137314 2020-01-23 09:22:58,215 Epoch: [466/484] Iter:[120/247], Time: 0.82, lr: [0.0005043009429133139], Loss: 0.136379 2020-01-23 09:23:06,193 Epoch: [466/484] Iter:[130/247], Time: 0.82, lr: [0.0005032516517629946], Loss: 0.135747 2020-01-23 09:23:14,257 Epoch: [466/484] Iter:[140/247], Time: 0.82, lr: [0.0005022021174677022], Loss: 0.136100 2020-01-23 09:23:22,528 Epoch: [466/484] Iter:[150/247], Time: 0.82, lr: [0.0005011523394062298], Loss: 0.135730 2020-01-23 09:23:30,543 Epoch: [466/484] Iter:[160/247], Time: 0.82, lr: [0.0005001023169543382], Loss: 0.134489 2020-01-23 09:23:38,673 Epoch: [466/484] Iter:[170/247], Time: 0.82, lr: [0.0004990520494847232], Loss: 0.134144 2020-01-23 09:23:46,771 Epoch: [466/484] Iter:[180/247], Time: 0.82, lr: [0.0004980015363670053], Loss: 0.133344 2020-01-23 09:23:54,991 Epoch: [466/484] Iter:[190/247], Time: 0.82, lr: [0.0004969507769676956], Loss: 0.132672 2020-01-23 09:24:03,383 Epoch: [466/484] Iter:[200/247], Time: 0.82, lr: [0.0004958997706501851], Loss: 0.131891 2020-01-23 09:24:11,542 Epoch: [466/484] Iter:[210/247], Time: 0.82, lr: [0.00049484851677471], Loss: 0.131304 2020-01-23 09:24:19,880 Epoch: [466/484] Iter:[220/247], Time: 0.82, lr: [0.00049379701469834], Loss: 0.130588 2020-01-23 09:24:27,994 Epoch: [466/484] Iter:[230/247], Time: 0.82, lr: [0.0004927452637749442], Loss: 0.131887 2020-01-23 09:24:36,314 Epoch: [466/484] Iter:[240/247], Time: 0.82, lr: [0.0004916932633551782], Loss: 0.132374 2020-01-23 09:28:05,268 0 [0.9840747 0.86766238 0.93204249 0.5585745 0.62785532 0.69651066 0.74480789 0.82135696 0.93049667 0.63864206 0.95202933 0.84037556 0.64049381 0.95243174 0.72367426 0.85690338 0.68002997 0.65783178 0.7940135 ] 0.7842003663235227 2020-01-23 09:28:05,269 1 [0.98426271 0.86906639 0.93211437 0.54639641 0.63016604 0.70047272 0.74596282 0.82275463 0.93097676 0.64439332 0.952163 0.84278969 0.64308678 0.95617205 0.79460638 0.87542434 0.73897895 0.66409791 0.79512237] 0.7931056656595974 2020-01-23 09:28:05,269 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 09:28:08,129 Loss: 0.166, MeanIU: 0.7931, Best_mIoU: 0.8033 2020-01-23 09:28:08,129 [0.98426271 0.86906639 0.93211437 0.54639641 0.63016604 0.70047272 0.74596282 0.82275463 0.93097676 0.64439332 0.952163 0.84278969 0.64308678 0.95617205 0.79460638 0.87542434 0.73897895 0.66409791 0.79512237] 2020-01-23 09:28:10,058 Epoch: [467/484] Iter:[0/247], Time: 1.92, lr: [0.0004909567142616703], Loss: 0.192269 2020-01-23 09:28:18,172 Epoch: [467/484] Iter:[10/247], Time: 0.91, lr: [0.0004899042881986302], Loss: 0.138270 2020-01-23 09:28:26,212 Epoch: [467/484] Iter:[20/247], Time: 0.86, lr: [0.0004888516108697188], Loss: 0.132952 2020-01-23 09:28:34,339 Epoch: [467/484] Iter:[30/247], Time: 0.85, lr: [0.00048779868161346885], Loss: 0.129351 2020-01-23 09:28:42,591 Epoch: [467/484] Iter:[40/247], Time: 0.84, lr: [0.00048674549976508594], Loss: 0.126491 2020-01-23 09:28:50,635 Epoch: [467/484] Iter:[50/247], Time: 0.83, lr: [0.00048569206465641333], Loss: 0.128602 2020-01-23 09:28:58,711 Epoch: [467/484] Iter:[60/247], Time: 0.83, lr: [0.0004846383756159175], Loss: 0.128057 2020-01-23 09:29:07,080 Epoch: [467/484] Iter:[70/247], Time: 0.83, lr: [0.00048358443196865167], Loss: 0.126289 2020-01-23 09:29:15,136 Epoch: [467/484] Iter:[80/247], Time: 0.83, lr: [0.00048253023303624146], Loss: 0.124461 2020-01-23 09:29:23,394 Epoch: [467/484] Iter:[90/247], Time: 0.83, lr: [0.00048147577813684807], Loss: 0.127203 2020-01-23 09:29:31,617 Epoch: [467/484] Iter:[100/247], Time: 0.83, lr: [0.00048042106658515295], Loss: 0.126757 2020-01-23 09:29:39,869 Epoch: [467/484] Iter:[110/247], Time: 0.83, lr: [0.0004793660976923206], Loss: 0.126285 2020-01-23 09:29:48,079 Epoch: [467/484] Iter:[120/247], Time: 0.83, lr: [0.000478310870765983], Loss: 0.126908 2020-01-23 09:29:56,093 Epoch: [467/484] Iter:[130/247], Time: 0.82, lr: [0.00047725538511020154], Loss: 0.127332 2020-01-23 09:30:04,325 Epoch: [467/484] Iter:[140/247], Time: 0.82, lr: [0.0004761996400254511], Loss: 0.125755 2020-01-23 09:30:12,494 Epoch: [467/484] Iter:[150/247], Time: 0.82, lr: [0.00047514363480858263], Loss: 0.125078 2020-01-23 09:30:20,656 Epoch: [467/484] Iter:[160/247], Time: 0.82, lr: [0.0004740873687527998], Loss: 0.124757 2020-01-23 09:30:29,137 Epoch: [467/484] Iter:[170/247], Time: 0.82, lr: [0.00047303084114763537], Loss: 0.125445 2020-01-23 09:30:37,204 Epoch: [467/484] Iter:[180/247], Time: 0.82, lr: [0.0004719740512789126], Loss: 0.125770 2020-01-23 09:30:45,344 Epoch: [467/484] Iter:[190/247], Time: 0.82, lr: [0.0004709169984287284], Loss: 0.126108 2020-01-23 09:30:53,466 Epoch: [467/484] Iter:[200/247], Time: 0.82, lr: [0.00046985968187541306], Loss: 0.126750 2020-01-23 09:31:01,569 Epoch: [467/484] Iter:[210/247], Time: 0.82, lr: [0.0004688021008935121], Loss: 0.127363 2020-01-23 09:31:09,605 Epoch: [467/484] Iter:[220/247], Time: 0.82, lr: [0.000467744254753746], Loss: 0.127027 2020-01-23 09:31:17,952 Epoch: [467/484] Iter:[230/247], Time: 0.82, lr: [0.0004666861427229913], Loss: 0.126537 2020-01-23 09:31:25,947 Epoch: [467/484] Iter:[240/247], Time: 0.82, lr: [0.0004656277640642396], Loss: 0.126481 2020-01-23 09:34:52,159 0 [0.98416989 0.86629179 0.93366442 0.56670372 0.63706748 0.69872507 0.74930258 0.82228465 0.92994666 0.62650546 0.95283151 0.84111052 0.64312305 0.95353196 0.73460485 0.83855002 0.6484054 0.6585812 0.7955607 ] 0.7832084711605984 2020-01-23 09:34:52,160 1 [0.98439409 0.86804416 0.93422364 0.56512522 0.64241215 0.70340037 0.75014413 0.82476729 0.93054254 0.62973297 0.95329735 0.84365237 0.65136345 0.95769721 0.81717694 0.85653287 0.67724067 0.66557657 0.79803111] 0.7922818481942623 2020-01-23 09:34:52,161 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 09:34:55,058 Loss: 0.163, MeanIU: 0.7923, Best_mIoU: 0.8033 2020-01-23 09:34:55,059 [0.98439409 0.86804416 0.93422364 0.56512522 0.64241215 0.70340037 0.75014413 0.82476729 0.93054254 0.62973297 0.95329735 0.84365237 0.65136345 0.95769721 0.81717694 0.85653287 0.67724067 0.66557657 0.79803111] 2020-01-23 09:34:57,040 Epoch: [468/484] Iter:[0/247], Time: 1.97, lr: [0.0004648867399628546], Loss: 0.095789 2020-01-23 09:35:05,080 Epoch: [468/484] Iter:[10/247], Time: 0.91, lr: [0.00046382790633400155], Loss: 0.121572 2020-01-23 09:35:13,392 Epoch: [468/484] Iter:[20/247], Time: 0.87, lr: [0.0004627688040676193], Loss: 0.129477 2020-01-23 09:35:21,430 Epoch: [468/484] Iter:[30/247], Time: 0.85, lr: [0.00046170943241208426], Loss: 0.130970 2020-01-23 09:35:29,519 Epoch: [468/484] Iter:[40/247], Time: 0.84, lr: [0.00046064979061174224], Loss: 0.129457 2020-01-23 09:35:37,795 Epoch: [468/484] Iter:[50/247], Time: 0.84, lr: [0.0004595898779068877], Loss: 0.128306 2020-01-23 09:35:45,783 Epoch: [468/484] Iter:[60/247], Time: 0.83, lr: [0.00045852969353372015], Loss: 0.127638 2020-01-23 09:35:53,944 Epoch: [468/484] Iter:[70/247], Time: 0.83, lr: [0.0004574692367243232], Loss: 0.125260 2020-01-23 09:36:02,080 Epoch: [468/484] Iter:[80/247], Time: 0.83, lr: [0.00045640850670662], Loss: 0.124835 2020-01-23 09:36:10,285 Epoch: [468/484] Iter:[90/247], Time: 0.83, lr: [0.00045534750270435144], Loss: 0.124889 2020-01-23 09:36:18,381 Epoch: [468/484] Iter:[100/247], Time: 0.82, lr: [0.00045428622393703133], Loss: 0.124268 2020-01-23 09:36:26,564 Epoch: [468/484] Iter:[110/247], Time: 0.82, lr: [0.0004532246696199234], Loss: 0.123612 2020-01-23 09:36:34,770 Epoch: [468/484] Iter:[120/247], Time: 0.82, lr: [0.0004521628389639959], Loss: 0.124026 2020-01-23 09:36:42,855 Epoch: [468/484] Iter:[130/247], Time: 0.82, lr: [0.0004511007311758982], Loss: 0.124232 2020-01-23 09:36:51,090 Epoch: [468/484] Iter:[140/247], Time: 0.82, lr: [0.0004500383454579143], Loss: 0.123487 2020-01-23 09:36:59,105 Epoch: [468/484] Iter:[150/247], Time: 0.82, lr: [0.00044897568100793885], Loss: 0.123524 2020-01-23 09:37:07,276 Epoch: [468/484] Iter:[160/247], Time: 0.82, lr: [0.00044791273701943107], Loss: 0.125154 2020-01-23 09:37:15,479 Epoch: [468/484] Iter:[170/247], Time: 0.82, lr: [0.0004468495126813831], Loss: 0.125463 2020-01-23 09:37:23,662 Epoch: [468/484] Iter:[180/247], Time: 0.82, lr: [0.0004457860071782876], Loss: 0.124663 2020-01-23 09:37:31,820 Epoch: [468/484] Iter:[190/247], Time: 0.82, lr: [0.0004447222196900905], Loss: 0.124511 2020-01-23 09:37:40,055 Epoch: [468/484] Iter:[200/247], Time: 0.82, lr: [0.0004436581493921653], Loss: 0.123212 2020-01-23 09:37:48,199 Epoch: [468/484] Iter:[210/247], Time: 0.82, lr: [0.00044259379545526375], Loss: 0.121892 2020-01-23 09:37:56,459 Epoch: [468/484] Iter:[220/247], Time: 0.82, lr: [0.0004415291570454889], Loss: 0.122300 2020-01-23 09:38:04,470 Epoch: [468/484] Iter:[230/247], Time: 0.82, lr: [0.0004404642333242449], Loss: 0.123231 2020-01-23 09:38:12,691 Epoch: [468/484] Iter:[240/247], Time: 0.82, lr: [0.00043939902344820994], Loss: 0.123620 2020-01-23 09:41:40,070 0 [0.98421468 0.86845627 0.93356008 0.57557084 0.63196503 0.69970147 0.74592588 0.82115901 0.93086626 0.63235585 0.95349201 0.83958856 0.64092599 0.95343701 0.7139013 0.84145848 0.65256003 0.65664666 0.79385064] 0.7826124244858883 2020-01-23 09:41:40,071 1 [0.98446138 0.87028255 0.93431881 0.58076283 0.63886388 0.70460688 0.7495364 0.82316477 0.9315974 0.63821583 0.953704 0.84172862 0.64653084 0.95708966 0.78195645 0.85717223 0.69414261 0.65971895 0.79568481] 0.7917652048397439 2020-01-23 09:41:40,071 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 09:41:42,869 Loss: 0.163, MeanIU: 0.7918, Best_mIoU: 0.8033 2020-01-23 09:41:42,870 [0.98446138 0.87028255 0.93431881 0.58076283 0.63886388 0.70460688 0.7495364 0.82316477 0.9315974 0.63821583 0.953704 0.84172862 0.64653084 0.95708966 0.78195645 0.85717223 0.69414261 0.65971895 0.79568481] 2020-01-23 09:41:44,925 Epoch: [469/484] Iter:[0/247], Time: 2.04, lr: [0.00043865320581891964], Loss: 0.124131 2020-01-23 09:41:53,181 Epoch: [469/484] Iter:[10/247], Time: 0.94, lr: [0.0004375875075307809], Loss: 0.129483 2020-01-23 09:42:01,434 Epoch: [469/484] Iter:[20/247], Time: 0.88, lr: [0.0004365215207869233], Loss: 0.117846 2020-01-23 09:42:09,949 Epoch: [469/484] Iter:[30/247], Time: 0.87, lr: [0.0004354552447261691], Loss: 0.122618 2020-01-23 09:42:18,125 Epoch: [469/484] Iter:[40/247], Time: 0.86, lr: [0.0004343886784824141], Loss: 0.121612 2020-01-23 09:42:26,318 Epoch: [469/484] Iter:[50/247], Time: 0.85, lr: [0.00043332182118459116], Loss: 0.123326 2020-01-23 09:42:34,476 Epoch: [469/484] Iter:[60/247], Time: 0.85, lr: [0.0004322546719566317], Loss: 0.122194 2020-01-23 09:42:42,754 Epoch: [469/484] Iter:[70/247], Time: 0.84, lr: [0.00043118722991741364], Loss: 0.118290 2020-01-23 09:42:51,020 Epoch: [469/484] Iter:[80/247], Time: 0.84, lr: [0.0004301194941807292], Loss: 0.122345 2020-01-23 09:42:59,164 Epoch: [469/484] Iter:[90/247], Time: 0.84, lr: [0.00042905146385523055], Loss: 0.123140 2020-01-23 09:43:07,232 Epoch: [469/484] Iter:[100/247], Time: 0.84, lr: [0.0004279831380443967], Loss: 0.124432 2020-01-23 09:43:15,586 Epoch: [469/484] Iter:[110/247], Time: 0.84, lr: [0.00042691451584647797], Loss: 0.124205 2020-01-23 09:43:23,843 Epoch: [469/484] Iter:[120/247], Time: 0.83, lr: [0.0004258455963544617], Loss: 0.123202 2020-01-23 09:43:31,918 Epoch: [469/484] Iter:[130/247], Time: 0.83, lr: [0.0004247763786560162], Loss: 0.124270 2020-01-23 09:43:40,011 Epoch: [469/484] Iter:[140/247], Time: 0.83, lr: [0.0004237068618334552], Loss: 0.124005 2020-01-23 09:43:48,374 Epoch: [469/484] Iter:[150/247], Time: 0.83, lr: [0.0004226370449636805], Loss: 0.124286 2020-01-23 09:43:56,642 Epoch: [469/484] Iter:[160/247], Time: 0.83, lr: [0.00042156692711814566], Loss: 0.123982 2020-01-23 09:44:04,793 Epoch: [469/484] Iter:[170/247], Time: 0.83, lr: [0.00042049650736279736], Loss: 0.125012 2020-01-23 09:44:13,059 Epoch: [469/484] Iter:[180/247], Time: 0.83, lr: [0.00041942578475803825], Loss: 0.124165 2020-01-23 09:44:21,159 Epoch: [469/484] Iter:[190/247], Time: 0.83, lr: [0.0004183547583586683], Loss: 0.124125 2020-01-23 09:44:29,213 Epoch: [469/484] Iter:[200/247], Time: 0.83, lr: [0.00041728342721383983], Loss: 0.124665 2020-01-23 09:44:37,374 Epoch: [469/484] Iter:[210/247], Time: 0.83, lr: [0.0004162117903670113], Loss: 0.123897 2020-01-23 09:44:45,449 Epoch: [469/484] Iter:[220/247], Time: 0.83, lr: [0.0004151398468558874], Loss: 0.123540 2020-01-23 09:44:53,469 Epoch: [469/484] Iter:[230/247], Time: 0.83, lr: [0.0004140675957123791], Loss: 0.122748 2020-01-23 09:45:01,845 Epoch: [469/484] Iter:[240/247], Time: 0.83, lr: [0.00041299503596254044], Loss: 0.124272 2020-01-23 09:48:19,328 0 [0.9840956 0.86730921 0.93317801 0.5633122 0.63841099 0.69804372 0.7436697 0.82326143 0.93083591 0.63634031 0.95149954 0.84099318 0.64325101 0.95351759 0.70881938 0.84018311 0.65365969 0.66097425 0.79424805] 0.7824001510075161 2020-01-23 09:48:19,329 1 [0.98434616 0.8693557 0.93396391 0.56419249 0.64354954 0.70298405 0.74559862 0.82613651 0.93162644 0.64149016 0.95190456 0.8439476 0.65122916 0.95593961 0.74250715 0.84880468 0.66641026 0.66937918 0.79625392] 0.7878747215533665 2020-01-23 09:48:19,330 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 09:48:22,259 Loss: 0.165, MeanIU: 0.7879, Best_mIoU: 0.8033 2020-01-23 09:48:22,259 [0.98434616 0.8693557 0.93396391 0.56419249 0.64354954 0.70298405 0.74559862 0.82613651 0.93162644 0.64149016 0.95190456 0.8439476 0.65122916 0.95593961 0.74250715 0.84880468 0.66641026 0.66937918 0.79625392] 2020-01-23 09:48:24,119 Epoch: [470/484] Iter:[0/247], Time: 1.85, lr: [0.00041224405999241927], Loss: 0.039987 2020-01-23 09:48:32,118 Epoch: [470/484] Iter:[10/247], Time: 0.90, lr: [0.00041117097335988535], Loss: 0.101448 2020-01-23 09:48:40,274 Epoch: [470/484] Iter:[20/247], Time: 0.86, lr: [0.00041009757546173865], Loss: 0.099401 2020-01-23 09:48:48,382 Epoch: [470/484] Iter:[30/247], Time: 0.84, lr: [0.0004090238653019236], Loss: 0.112577 2020-01-23 09:48:56,641 Epoch: [470/484] Iter:[40/247], Time: 0.84, lr: [0.0004079498418782876], Loss: 0.117014 2020-01-23 09:49:05,056 Epoch: [470/484] Iter:[50/247], Time: 0.84, lr: [0.00040687550418251546], Loss: 0.121828 2020-01-23 09:49:13,139 Epoch: [470/484] Iter:[60/247], Time: 0.83, lr: [0.00040580085120007803], Loss: 0.122947 2020-01-23 09:49:21,139 Epoch: [470/484] Iter:[70/247], Time: 0.83, lr: [0.00040472588191017963], Loss: 0.125027 2020-01-23 09:49:29,464 Epoch: [470/484] Iter:[80/247], Time: 0.83, lr: [0.0004036505952856905], Loss: 0.125593 2020-01-23 09:49:37,760 Epoch: [470/484] Iter:[90/247], Time: 0.83, lr: [0.00040257499029310034], Loss: 0.127033 2020-01-23 09:49:45,916 Epoch: [470/484] Iter:[100/247], Time: 0.83, lr: [0.0004014990658924481], Loss: 0.126176 2020-01-23 09:49:54,072 Epoch: [470/484] Iter:[110/247], Time: 0.83, lr: [0.0004004228210372738], Loss: 0.123730 2020-01-23 09:50:02,149 Epoch: [470/484] Iter:[120/247], Time: 0.83, lr: [0.0003993462546745468], Loss: 0.124894 2020-01-23 09:50:10,191 Epoch: [470/484] Iter:[130/247], Time: 0.82, lr: [0.00039826936574461654], Loss: 0.124453 2020-01-23 09:50:18,348 Epoch: [470/484] Iter:[140/247], Time: 0.82, lr: [0.000397192153181139], Loss: 0.124200 2020-01-23 09:50:26,500 Epoch: [470/484] Iter:[150/247], Time: 0.82, lr: [0.000396114615911026], Loss: 0.125895 2020-01-23 09:50:34,658 Epoch: [470/484] Iter:[160/247], Time: 0.82, lr: [0.0003950367528543703], Loss: 0.125300 2020-01-23 09:50:42,794 Epoch: [470/484] Iter:[170/247], Time: 0.82, lr: [0.0003939585629243929], Loss: 0.124929 2020-01-23 09:50:50,876 Epoch: [470/484] Iter:[180/247], Time: 0.82, lr: [0.000392880045027367], Loss: 0.124607 2020-01-23 09:50:59,084 Epoch: [470/484] Iter:[190/247], Time: 0.82, lr: [0.0003918011980625637], Loss: 0.124130 2020-01-23 09:51:07,237 Epoch: [470/484] Iter:[200/247], Time: 0.82, lr: [0.0003907220209221751], Loss: 0.123993 2020-01-23 09:51:15,326 Epoch: [470/484] Iter:[210/247], Time: 0.82, lr: [0.0003896425124912521], Loss: 0.124741 2020-01-23 09:51:23,571 Epoch: [470/484] Iter:[220/247], Time: 0.82, lr: [0.00038856267164763975], Loss: 0.124787 2020-01-23 09:51:31,754 Epoch: [470/484] Iter:[230/247], Time: 0.82, lr: [0.0003874824972618985], Loss: 0.125405 2020-01-23 09:51:39,980 Epoch: [470/484] Iter:[240/247], Time: 0.82, lr: [0.0003864019881972457], Loss: 0.125794 2020-01-23 09:55:01,406 0 [0.98423083 0.86923941 0.9338234 0.5920862 0.63495497 0.69995981 0.74640283 0.81986434 0.9307016 0.64054152 0.95318395 0.84021291 0.64032014 0.9536651 0.69531711 0.83462909 0.65010572 0.65491831 0.79172293] 0.7824147460175406 2020-01-23 09:55:01,407 1 [0.98442465 0.87087922 0.93439038 0.59090588 0.6396356 0.7039103 0.74763751 0.82252763 0.93115218 0.64524046 0.95353651 0.8422542 0.64864212 0.95758829 0.75060984 0.83646595 0.65975538 0.66528995 0.79489851] 0.7884076086552868 2020-01-23 09:55:01,407 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 09:55:04,316 Loss: 0.164, MeanIU: 0.7884, Best_mIoU: 0.8033 2020-01-23 09:55:04,317 [0.98442465 0.87087922 0.93439038 0.59090588 0.6396356 0.7039103 0.74763751 0.82252763 0.93115218 0.64524046 0.95353651 0.8422542 0.64864212 0.95758829 0.75060984 0.83646595 0.65975538 0.66528995 0.79489851] 2020-01-23 09:55:06,143 Epoch: [471/484] Iter:[0/247], Time: 1.82, lr: [0.0003856454321056124], Loss: 0.180211 2020-01-23 09:55:14,238 Epoch: [471/484] Iter:[10/247], Time: 0.90, lr: [0.0003845643514568483], Loss: 0.115277 2020-01-23 09:55:22,287 Epoch: [471/484] Iter:[20/247], Time: 0.86, lr: [0.00038348293302258463], Loss: 0.118951 2020-01-23 09:55:30,415 Epoch: [471/484] Iter:[30/247], Time: 0.84, lr: [0.0003824011756382254], Loss: 0.118626 2020-01-23 09:55:38,900 Epoch: [471/484] Iter:[40/247], Time: 0.84, lr: [0.0003813190781314795], Loss: 0.122733 2020-01-23 09:55:47,192 Epoch: [471/484] Iter:[50/247], Time: 0.84, lr: [0.0003802366393222965], Loss: 0.122352 2020-01-23 09:55:55,524 Epoch: [471/484] Iter:[60/247], Time: 0.84, lr: [0.00037915385802277937], Loss: 0.127918 2020-01-23 09:56:03,610 Epoch: [471/484] Iter:[70/247], Time: 0.84, lr: [0.00037807073303711884], Loss: 0.125376 2020-01-23 09:56:11,951 Epoch: [471/484] Iter:[80/247], Time: 0.83, lr: [0.0003769872631615052], Loss: 0.125997 2020-01-23 09:56:20,026 Epoch: [471/484] Iter:[90/247], Time: 0.83, lr: [0.000375903447184053], Loss: 0.126431 2020-01-23 09:56:28,128 Epoch: [471/484] Iter:[100/247], Time: 0.83, lr: [0.0003748192838847258], Loss: 0.125811 2020-01-23 09:56:36,215 Epoch: [471/484] Iter:[110/247], Time: 0.83, lr: [0.00037373477203524413], Loss: 0.126298 2020-01-23 09:56:44,370 Epoch: [471/484] Iter:[120/247], Time: 0.83, lr: [0.00037264991039901536], Loss: 0.126510 2020-01-23 09:56:52,447 Epoch: [471/484] Iter:[130/247], Time: 0.83, lr: [0.00037156469773103816], Loss: 0.125788 2020-01-23 09:57:00,601 Epoch: [471/484] Iter:[140/247], Time: 0.82, lr: [0.00037047913277782964], Loss: 0.124933 2020-01-23 09:57:08,690 Epoch: [471/484] Iter:[150/247], Time: 0.82, lr: [0.0003693932142773283], Loss: 0.124703 2020-01-23 09:57:16,653 Epoch: [471/484] Iter:[160/247], Time: 0.82, lr: [0.0003683069409588184], Loss: 0.123877 2020-01-23 09:57:24,789 Epoch: [471/484] Iter:[170/247], Time: 0.82, lr: [0.00036722031154283084], Loss: 0.124481 2020-01-23 09:57:32,806 Epoch: [471/484] Iter:[180/247], Time: 0.82, lr: [0.0003661333247410652], Loss: 0.124005 2020-01-23 09:57:40,978 Epoch: [471/484] Iter:[190/247], Time: 0.82, lr: [0.0003650459792562877], Loss: 0.122907 2020-01-23 09:57:48,997 Epoch: [471/484] Iter:[200/247], Time: 0.82, lr: [0.00036395827378225154], Loss: 0.121771 2020-01-23 09:57:57,032 Epoch: [471/484] Iter:[210/247], Time: 0.82, lr: [0.0003628702070035921], Loss: 0.121584 2020-01-23 09:58:05,128 Epoch: [471/484] Iter:[220/247], Time: 0.82, lr: [0.00036178177759574456], Loss: 0.120766 2020-01-23 09:58:13,268 Epoch: [471/484] Iter:[230/247], Time: 0.82, lr: [0.0003606929842248383], Loss: 0.121544 2020-01-23 09:58:21,503 Epoch: [471/484] Iter:[240/247], Time: 0.82, lr: [0.0003596038255476044], Loss: 0.121867 2020-01-23 10:01:57,899 0 [0.98439247 0.86920535 0.93321118 0.56817043 0.63291181 0.6994973 0.74555977 0.82047236 0.93020934 0.63934399 0.95246153 0.84112153 0.64371005 0.95402116 0.72247761 0.85736517 0.7106087 0.65463054 0.79433955] 0.7870373599835038 2020-01-23 10:01:57,900 1 [0.98461203 0.87093956 0.93366366 0.55956686 0.63759038 0.70347689 0.74711618 0.82292895 0.93087278 0.64416221 0.95310819 0.84338623 0.64962671 0.95824935 0.78614686 0.86325307 0.74709756 0.65956291 0.79566469] 0.7942644778143952 2020-01-23 10:01:57,900 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 10:02:00,844 Loss: 0.163, MeanIU: 0.7943, Best_mIoU: 0.8033 2020-01-23 10:02:00,844 [0.98461203 0.87093956 0.93366366 0.55956686 0.63759038 0.70347689 0.74711618 0.82292895 0.93087278 0.64416221 0.95310819 0.84338623 0.64962671 0.95824935 0.78614686 0.86325307 0.74709756 0.65956291 0.79566469] 2020-01-23 10:02:02,755 Epoch: [472/484] Iter:[0/247], Time: 1.90, lr: [0.00035884119639225694], Loss: 0.067849 2020-01-23 10:02:10,893 Epoch: [472/484] Iter:[10/247], Time: 0.91, lr: [0.00035775141358454125], Loss: 0.120018 2020-01-23 10:02:19,131 Epoch: [472/484] Iter:[20/247], Time: 0.87, lr: [0.0003566612617959625], Loss: 0.117600 2020-01-23 10:02:27,173 Epoch: [472/484] Iter:[30/247], Time: 0.85, lr: [0.00035557073964761343], Loss: 0.119953 2020-01-23 10:02:35,319 Epoch: [472/484] Iter:[40/247], Time: 0.84, lr: [0.0003544798457507213], Loss: 0.127204 2020-01-23 10:02:43,482 Epoch: [472/484] Iter:[50/247], Time: 0.84, lr: [0.00035338857870653197], Loss: 0.127339 2020-01-23 10:02:51,684 Epoch: [472/484] Iter:[60/247], Time: 0.83, lr: [0.00035229693710621466], Loss: 0.125563 2020-01-23 10:02:59,749 Epoch: [472/484] Iter:[70/247], Time: 0.83, lr: [0.00035120491953074347], Loss: 0.126410 2020-01-23 10:03:07,790 Epoch: [472/484] Iter:[80/247], Time: 0.83, lr: [0.00035011252455079925], Loss: 0.125490 2020-01-23 10:03:15,956 Epoch: [472/484] Iter:[90/247], Time: 0.83, lr: [0.0003490197507266491], Loss: 0.126361 2020-01-23 10:03:24,147 Epoch: [472/484] Iter:[100/247], Time: 0.82, lr: [0.0003479265966080382], Loss: 0.126115 2020-01-23 10:03:32,210 Epoch: [472/484] Iter:[110/247], Time: 0.82, lr: [0.00034683306073408037], Loss: 0.126041 2020-01-23 10:03:40,412 Epoch: [472/484] Iter:[120/247], Time: 0.82, lr: [0.0003457391416331322], Loss: 0.127136 2020-01-23 10:03:48,545 Epoch: [472/484] Iter:[130/247], Time: 0.82, lr: [0.0003446448378226878], Loss: 0.127279 2020-01-23 10:03:56,671 Epoch: [472/484] Iter:[140/247], Time: 0.82, lr: [0.0003435501478092477], Loss: 0.126418 2020-01-23 10:04:04,857 Epoch: [472/484] Iter:[150/247], Time: 0.82, lr: [0.00034245507008821054], Loss: 0.125961 2020-01-23 10:04:12,798 Epoch: [472/484] Iter:[160/247], Time: 0.82, lr: [0.00034135960314373847], Loss: 0.124526 2020-01-23 10:04:20,850 Epoch: [472/484] Iter:[170/247], Time: 0.82, lr: [0.00034026374544864486], Loss: 0.125002 2020-01-23 10:04:28,854 Epoch: [472/484] Iter:[180/247], Time: 0.82, lr: [0.00033916749546425656], Loss: 0.124476 2020-01-23 10:04:36,917 Epoch: [472/484] Iter:[190/247], Time: 0.82, lr: [0.000338070851640298], Loss: 0.123885 2020-01-23 10:04:44,955 Epoch: [472/484] Iter:[200/247], Time: 0.82, lr: [0.0003369738124147491], Loss: 0.123679 2020-01-23 10:04:53,102 Epoch: [472/484] Iter:[210/247], Time: 0.82, lr: [0.00033587637621372595], Loss: 0.124270 2020-01-23 10:05:01,183 Epoch: [472/484] Iter:[220/247], Time: 0.82, lr: [0.00033477854145133525], Loss: 0.124459 2020-01-23 10:05:09,449 Epoch: [472/484] Iter:[230/247], Time: 0.82, lr: [0.0003336803065295503], Loss: 0.124599 2020-01-23 10:05:17,707 Epoch: [472/484] Iter:[240/247], Time: 0.82, lr: [0.000332581669838063], Loss: 0.124362 2020-01-23 10:08:44,819 0 [0.98460027 0.87121124 0.93284538 0.56538873 0.6322862 0.69790457 0.74697931 0.82409515 0.9306186 0.63789396 0.95313121 0.83652203 0.61298419 0.95376341 0.71209127 0.85860158 0.72032112 0.66003605 0.79459724] 0.7855721847231726 2020-01-23 10:08:44,820 1 [0.98483833 0.87301399 0.9329803 0.55671737 0.63555377 0.70142407 0.74969474 0.82617042 0.93118896 0.64180866 0.95405021 0.8402508 0.6261658 0.95734371 0.76971672 0.86825007 0.75508706 0.6625382 0.79520116] 0.7927365447742296 2020-01-23 10:08:44,821 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 10:08:47,615 Loss: 0.165, MeanIU: 0.7927, Best_mIoU: 0.8033 2020-01-23 10:08:47,616 [0.98483833 0.87301399 0.9329803 0.55671737 0.63555377 0.70142407 0.74969474 0.82617042 0.93118896 0.64180866 0.95405021 0.8402508 0.6261658 0.95734371 0.76971672 0.86825007 0.75508706 0.6625382 0.79520116] 2020-01-23 10:08:49,553 Epoch: [473/484] Iter:[0/247], Time: 1.93, lr: [0.00033181238423258223], Loss: 0.168255 2020-01-23 10:08:57,576 Epoch: [473/484] Iter:[10/247], Time: 0.90, lr: [0.0003307130608017321], Loss: 0.136556 2020-01-23 10:09:05,545 Epoch: [473/484] Iter:[20/247], Time: 0.85, lr: [0.00032961333119092704], Loss: 0.126938 2020-01-23 10:09:13,604 Epoch: [473/484] Iter:[30/247], Time: 0.84, lr: [0.0003285131937432072], Loss: 0.127144 2020-01-23 10:09:21,698 Epoch: [473/484] Iter:[40/247], Time: 0.83, lr: [0.0003274126467886547], Loss: 0.123490 2020-01-23 10:09:29,696 Epoch: [473/484] Iter:[50/247], Time: 0.82, lr: [0.00032631168864425504], Loss: 0.121994 2020-01-23 10:09:37,750 Epoch: [473/484] Iter:[60/247], Time: 0.82, lr: [0.00032521031761373326], Loss: 0.123384 2020-01-23 10:09:46,001 Epoch: [473/484] Iter:[70/247], Time: 0.82, lr: [0.0003241085319874109], Loss: 0.124742 2020-01-23 10:09:54,055 Epoch: [473/484] Iter:[80/247], Time: 0.82, lr: [0.0003230063300420371], Loss: 0.124162 2020-01-23 10:10:02,117 Epoch: [473/484] Iter:[90/247], Time: 0.82, lr: [0.00032190371004064104], Loss: 0.123241 2020-01-23 10:10:10,166 Epoch: [473/484] Iter:[100/247], Time: 0.82, lr: [0.00032080067023235957], Loss: 0.124651 2020-01-23 10:10:18,450 Epoch: [473/484] Iter:[110/247], Time: 0.82, lr: [0.0003196972088522769], Loss: 0.125001 2020-01-23 10:10:26,422 Epoch: [473/484] Iter:[120/247], Time: 0.82, lr: [0.00031859332412126216], Loss: 0.125901 2020-01-23 10:10:34,529 Epoch: [473/484] Iter:[130/247], Time: 0.82, lr: [0.000317489014245789], Loss: 0.124395 2020-01-23 10:10:42,751 Epoch: [473/484] Iter:[140/247], Time: 0.82, lr: [0.0003163842774177754], Loss: 0.126242 2020-01-23 10:10:50,906 Epoch: [473/484] Iter:[150/247], Time: 0.82, lr: [0.0003152791118143961], Loss: 0.126225 2020-01-23 10:10:59,136 Epoch: [473/484] Iter:[160/247], Time: 0.82, lr: [0.00031417351559791674], Loss: 0.125683 2020-01-23 10:11:07,174 Epoch: [473/484] Iter:[170/247], Time: 0.82, lr: [0.0003130674869155011], Loss: 0.124949 2020-01-23 10:11:15,501 Epoch: [473/484] Iter:[180/247], Time: 0.82, lr: [0.0003119610238990393], Loss: 0.124872 2020-01-23 10:11:23,686 Epoch: [473/484] Iter:[190/247], Time: 0.82, lr: [0.00031085412466494894], Loss: 0.124061 2020-01-23 10:11:31,924 Epoch: [473/484] Iter:[200/247], Time: 0.82, lr: [0.0003097467873139973], Loss: 0.125194 2020-01-23 10:11:40,290 Epoch: [473/484] Iter:[210/247], Time: 0.82, lr: [0.00030863900993109673], Loss: 0.125132 2020-01-23 10:11:48,655 Epoch: [473/484] Iter:[220/247], Time: 0.82, lr: [0.0003075307905851203], Loss: 0.125381 2020-01-23 10:11:56,758 Epoch: [473/484] Iter:[230/247], Time: 0.82, lr: [0.0003064221273286908], Loss: 0.125238 2020-01-23 10:12:05,042 Epoch: [473/484] Iter:[240/247], Time: 0.82, lr: [0.0003053130181979902], Loss: 0.125287 2020-01-23 10:15:30,148 0 [0.98461069 0.87106202 0.93403975 0.58791253 0.63530438 0.70057261 0.74770854 0.81851507 0.93086016 0.63817573 0.95280532 0.83905178 0.63607439 0.95365741 0.70782817 0.85417728 0.69421426 0.65094056 0.79167975] 0.7857468623524323 2020-01-23 10:15:30,149 1 [0.984774 0.87270805 0.93499873 0.59689198 0.64139304 0.70496067 0.74979725 0.82123185 0.93150431 0.64110247 0.95309993 0.84119897 0.64154716 0.9574447 0.76191701 0.87036539 0.77654558 0.65550062 0.79260419] 0.7962939946868476 2020-01-23 10:15:30,150 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 10:15:33,082 Loss: 0.162, MeanIU: 0.7963, Best_mIoU: 0.8033 2020-01-23 10:15:33,083 [0.984774 0.87270805 0.93499873 0.59689198 0.64139304 0.70496067 0.74979725 0.82123185 0.93150431 0.64110247 0.95309993 0.84119897 0.64154716 0.9574447 0.76191701 0.87036539 0.77654558 0.65550062 0.79260419] 2020-01-23 10:15:34,951 Epoch: [474/484] Iter:[0/247], Time: 1.86, lr: [0.0003045363754514388], Loss: 0.121844 2020-01-23 10:15:43,010 Epoch: [474/484] Iter:[10/247], Time: 0.90, lr: [0.00030342650378026455], Loss: 0.137934 2020-01-23 10:15:51,200 Epoch: [474/484] Iter:[20/247], Time: 0.86, lr: [0.00030231618084872744], Loss: 0.140091 2020-01-23 10:15:59,366 Epoch: [474/484] Iter:[30/247], Time: 0.85, lr: [0.0003012054046303328], Loss: 0.138707 2020-01-23 10:16:07,465 Epoch: [474/484] Iter:[40/247], Time: 0.84, lr: [0.000300094173081147], Loss: 0.137068 2020-01-23 10:16:15,475 Epoch: [474/484] Iter:[50/247], Time: 0.83, lr: [0.0002989824841395632], Loss: 0.132649 2020-01-23 10:16:23,504 Epoch: [474/484] Iter:[60/247], Time: 0.83, lr: [0.0002978703357260865], Loss: 0.129489 2020-01-23 10:16:31,516 Epoch: [474/484] Iter:[70/247], Time: 0.82, lr: [0.0002967577257430922], Loss: 0.127531 2020-01-23 10:16:39,832 Epoch: [474/484] Iter:[80/247], Time: 0.82, lr: [0.0002956446520746032], Loss: 0.125706 2020-01-23 10:16:48,128 Epoch: [474/484] Iter:[90/247], Time: 0.82, lr: [0.00029453111258604027], Loss: 0.125463 2020-01-23 10:16:56,323 Epoch: [474/484] Iter:[100/247], Time: 0.82, lr: [0.0002934171051239915], Loss: 0.123881 2020-01-23 10:17:04,526 Epoch: [474/484] Iter:[110/247], Time: 0.82, lr: [0.0002923026275159539], Loss: 0.125373 2020-01-23 10:17:12,548 Epoch: [474/484] Iter:[120/247], Time: 0.82, lr: [0.00029118767757009453], Loss: 0.124109 2020-01-23 10:17:20,692 Epoch: [474/484] Iter:[130/247], Time: 0.82, lr: [0.0002900722530749851], Loss: 0.124937 2020-01-23 10:17:28,952 Epoch: [474/484] Iter:[140/247], Time: 0.82, lr: [0.0002889563517993469], Loss: 0.123825 2020-01-23 10:17:37,065 Epoch: [474/484] Iter:[150/247], Time: 0.82, lr: [0.0002878399714917906], Loss: 0.124250 2020-01-23 10:17:45,206 Epoch: [474/484] Iter:[160/247], Time: 0.82, lr: [0.0002867231098805378], Loss: 0.124178 2020-01-23 10:17:53,459 Epoch: [474/484] Iter:[170/247], Time: 0.82, lr: [0.0002856057646731586], Loss: 0.123486 2020-01-23 10:18:01,760 Epoch: [474/484] Iter:[180/247], Time: 0.82, lr: [0.0002844879335562814], Loss: 0.122757 2020-01-23 10:18:09,834 Epoch: [474/484] Iter:[190/247], Time: 0.82, lr: [0.00028336961419532183], Loss: 0.123733 2020-01-23 10:18:17,891 Epoch: [474/484] Iter:[200/247], Time: 0.82, lr: [0.0002822508042341815], Loss: 0.122272 2020-01-23 10:18:25,973 Epoch: [474/484] Iter:[210/247], Time: 0.82, lr: [0.0002811315012949666], Loss: 0.122155 2020-01-23 10:18:34,077 Epoch: [474/484] Iter:[220/247], Time: 0.82, lr: [0.00028001170297767676], Loss: 0.122021 2020-01-23 10:18:42,207 Epoch: [474/484] Iter:[230/247], Time: 0.82, lr: [0.0002788914068599125], Loss: 0.122613 2020-01-23 10:18:50,289 Epoch: [474/484] Iter:[240/247], Time: 0.82, lr: [0.000277770610496553], Loss: 0.122502 2020-01-23 10:22:13,579 0 [0.98438447 0.8689603 0.93394711 0.58248262 0.6401366 0.70126636 0.74894927 0.82460367 0.93029219 0.63474259 0.95337494 0.83744663 0.61559994 0.95295242 0.69671058 0.86105973 0.73064992 0.66129334 0.79356086] 0.7869691335670714 2020-01-23 10:22:13,580 1 [0.98459176 0.87064074 0.93480229 0.58548867 0.64555054 0.70544607 0.75069015 0.82760819 0.93100747 0.63983786 0.95384572 0.84050926 0.62251865 0.9565583 0.75003689 0.87566674 0.79792343 0.67013574 0.79541314] 0.7967511376250825 2020-01-23 10:22:13,580 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 10:22:16,405 Loss: 0.163, MeanIU: 0.7968, Best_mIoU: 0.8033 2020-01-23 10:22:16,405 [0.98459176 0.87064074 0.93480229 0.58548867 0.64555054 0.70544607 0.75069015 0.82760819 0.93100747 0.63983786 0.95384572 0.84050926 0.62251865 0.9565583 0.75003689 0.87566674 0.79792343 0.67013574 0.79541314] 2020-01-23 10:22:18,327 Epoch: [475/484] Iter:[0/247], Time: 1.91, lr: [0.0002769857540753147], Loss: 0.144831 2020-01-23 10:22:26,527 Epoch: [475/484] Iter:[10/247], Time: 0.92, lr: [0.00027586410161755277], Loss: 0.134913 2020-01-23 10:22:34,720 Epoch: [475/484] Iter:[20/247], Time: 0.87, lr: [0.0002747419421964232], Loss: 0.123261 2020-01-23 10:22:43,061 Epoch: [475/484] Iter:[30/247], Time: 0.86, lr: [0.000273619273279972], Loss: 0.124412 2020-01-23 10:22:51,270 Epoch: [475/484] Iter:[40/247], Time: 0.85, lr: [0.00027249609231198765], Loss: 0.121076 2020-01-23 10:22:59,384 Epoch: [475/484] Iter:[50/247], Time: 0.84, lr: [0.00027137239671166854], Loss: 0.123613 2020-01-23 10:23:07,420 Epoch: [475/484] Iter:[60/247], Time: 0.84, lr: [0.00027024818387326], Loss: 0.127117 2020-01-23 10:23:15,528 Epoch: [475/484] Iter:[70/247], Time: 0.83, lr: [0.0002691234511657087], Loss: 0.124037 2020-01-23 10:23:23,841 Epoch: [475/484] Iter:[80/247], Time: 0.83, lr: [0.00026799819593228616], Loss: 0.122026 2020-01-23 10:23:32,000 Epoch: [475/484] Iter:[90/247], Time: 0.83, lr: [0.00026687241549022944], Loss: 0.120724 2020-01-23 10:23:40,088 Epoch: [475/484] Iter:[100/247], Time: 0.83, lr: [0.0002657461071303505], Loss: 0.120797 2020-01-23 10:23:48,203 Epoch: [475/484] Iter:[110/247], Time: 0.83, lr: [0.0002646192681166624], Loss: 0.122866 2020-01-23 10:23:56,300 Epoch: [475/484] Iter:[120/247], Time: 0.83, lr: [0.0002634918956859737], Loss: 0.123772 2020-01-23 10:24:04,533 Epoch: [475/484] Iter:[130/247], Time: 0.83, lr: [0.00026236398704749983], Loss: 0.123734 2020-01-23 10:24:12,739 Epoch: [475/484] Iter:[140/247], Time: 0.82, lr: [0.0002612355393824434], Loss: 0.123392 2020-01-23 10:24:20,982 Epoch: [475/484] Iter:[150/247], Time: 0.82, lr: [0.000260106549843582], Loss: 0.123441 2020-01-23 10:24:29,129 Epoch: [475/484] Iter:[160/247], Time: 0.82, lr: [0.00025897701555484715], Loss: 0.123205 2020-01-23 10:24:37,395 Epoch: [475/484] Iter:[170/247], Time: 0.82, lr: [0.00025784693361088066], Loss: 0.123151 2020-01-23 10:24:45,584 Epoch: [475/484] Iter:[180/247], Time: 0.82, lr: [0.0002567163010766038], Loss: 0.123639 2020-01-23 10:24:53,826 Epoch: [475/484] Iter:[190/247], Time: 0.82, lr: [0.00025558511498675467], Loss: 0.123323 2020-01-23 10:25:01,932 Epoch: [475/484] Iter:[200/247], Time: 0.82, lr: [0.0002544533723454395], Loss: 0.121954 2020-01-23 10:25:10,267 Epoch: [475/484] Iter:[210/247], Time: 0.82, lr: [0.00025332107012565115], Loss: 0.122238 2020-01-23 10:25:18,401 Epoch: [475/484] Iter:[220/247], Time: 0.82, lr: [0.00025218820526880204], Loss: 0.122532 2020-01-23 10:25:26,524 Epoch: [475/484] Iter:[230/247], Time: 0.82, lr: [0.00025105477468422303], Loss: 0.122711 2020-01-23 10:25:34,581 Epoch: [475/484] Iter:[240/247], Time: 0.82, lr: [0.00024992077524867655], Loss: 0.121795 2020-01-23 10:29:06,466 0 [0.98408551 0.868231 0.93364447 0.57803482 0.64003772 0.70113068 0.74795124 0.82561553 0.9300227 0.63713769 0.95219838 0.84044121 0.63178273 0.95319815 0.70045566 0.86083376 0.73215374 0.6619401 0.79457411] 0.7880773267445321 2020-01-23 10:29:06,467 1 [0.98431864 0.86987782 0.93438542 0.58084935 0.64438789 0.70528434 0.74978093 0.82764136 0.93061144 0.6400203 0.95349152 0.84289365 0.6376478 0.9564287 0.74820027 0.87484201 0.79381354 0.66974945 0.79585455] 0.796846262096388 2020-01-23 10:29:06,467 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 10:29:09,277 Loss: 0.164, MeanIU: 0.7968, Best_mIoU: 0.8033 2020-01-23 10:29:09,277 [0.98431864 0.86987782 0.93438542 0.58084935 0.64438789 0.70528434 0.74978093 0.82764136 0.93061144 0.6400203 0.95349152 0.84289365 0.6376478 0.9564287 0.74820027 0.87484201 0.79381354 0.66974945 0.79585455] 2020-01-23 10:29:11,265 Epoch: [476/484] Iter:[0/247], Time: 1.98, lr: [0.0002491266354885982], Loss: 0.100399 2020-01-23 10:29:19,495 Epoch: [476/484] Iter:[10/247], Time: 0.93, lr: [0.0002479916617446659], Loss: 0.122167 2020-01-23 10:29:27,631 Epoch: [476/484] Iter:[20/247], Time: 0.87, lr: [0.00024685611055028344], Loss: 0.116266 2020-01-23 10:29:35,712 Epoch: [476/484] Iter:[30/247], Time: 0.85, lr: [0.0002457199786571703], Loss: 0.119998 2020-01-23 10:29:43,810 Epoch: [476/484] Iter:[40/247], Time: 0.84, lr: [0.0002445832627819873], Loss: 0.125454 2020-01-23 10:29:51,864 Epoch: [476/484] Iter:[50/247], Time: 0.83, lr: [0.00024344595960576498], Loss: 0.124229 2020-01-23 10:29:59,963 Epoch: [476/484] Iter:[60/247], Time: 0.83, lr: [0.00024230806577334185], Loss: 0.121739 2020-01-23 10:30:08,006 Epoch: [476/484] Iter:[70/247], Time: 0.83, lr: [0.00024116957789276666], Loss: 0.121234 2020-01-23 10:30:16,346 Epoch: [476/484] Iter:[80/247], Time: 0.83, lr: [0.00024003049253471185], Loss: 0.120092 2020-01-23 10:30:24,519 Epoch: [476/484] Iter:[90/247], Time: 0.83, lr: [0.00023889080623185028], Loss: 0.118837 2020-01-23 10:30:32,713 Epoch: [476/484] Iter:[100/247], Time: 0.83, lr: [0.0002377505154782425], Loss: 0.120069 2020-01-23 10:30:40,878 Epoch: [476/484] Iter:[110/247], Time: 0.83, lr: [0.00023660961672868641], Loss: 0.119016 2020-01-23 10:30:48,919 Epoch: [476/484] Iter:[120/247], Time: 0.82, lr: [0.00023546810639807692], Loss: 0.119222 2020-01-23 10:30:57,103 Epoch: [476/484] Iter:[130/247], Time: 0.82, lr: [0.00023432598086072718], Loss: 0.118169 2020-01-23 10:31:05,113 Epoch: [476/484] Iter:[140/247], Time: 0.82, lr: [0.000233183236449699], Loss: 0.117266 2020-01-23 10:31:13,231 Epoch: [476/484] Iter:[150/247], Time: 0.82, lr: [0.000232039869456094], Loss: 0.117394 2020-01-23 10:31:21,490 Epoch: [476/484] Iter:[160/247], Time: 0.82, lr: [0.00023089587612835317], Loss: 0.116930 2020-01-23 10:31:29,653 Epoch: [476/484] Iter:[170/247], Time: 0.82, lr: [0.00022975125267151776], Loss: 0.117161 2020-01-23 10:31:37,922 Epoch: [476/484] Iter:[180/247], Time: 0.82, lr: [0.0002286059952464888], Loss: 0.116967 2020-01-23 10:31:46,183 Epoch: [476/484] Iter:[190/247], Time: 0.82, lr: [0.00022746009996926943], Loss: 0.117128 2020-01-23 10:31:54,496 Epoch: [476/484] Iter:[200/247], Time: 0.82, lr: [0.0002263135629101747], Loss: 0.117791 2020-01-23 10:32:02,737 Epoch: [476/484] Iter:[210/247], Time: 0.82, lr: [0.00022516638009304605], Loss: 0.119141 2020-01-23 10:32:11,023 Epoch: [476/484] Iter:[220/247], Time: 0.82, lr: [0.0002240185474944231], Loss: 0.118622 2020-01-23 10:32:19,317 Epoch: [476/484] Iter:[230/247], Time: 0.82, lr: [0.00022287006104272061], Loss: 0.117878 2020-01-23 10:32:27,507 Epoch: [476/484] Iter:[240/247], Time: 0.82, lr: [0.0002217209166173614], Loss: 0.118172 2020-01-23 10:35:54,122 0 [0.98418462 0.86917879 0.93391301 0.57976346 0.63798334 0.70298022 0.75036525 0.8237418 0.93029237 0.63976399 0.95244916 0.84050093 0.63465386 0.9536618 0.71471392 0.86069193 0.71138766 0.66735087 0.79616166] 0.7886178234357949 2020-01-23 10:35:54,123 1 [0.98436086 0.87094216 0.93478097 0.58798563 0.64561136 0.70716943 0.75270548 0.82637526 0.93108823 0.64316581 0.95323301 0.84296612 0.63817268 0.95687612 0.76192188 0.87704648 0.77750263 0.67032026 0.79737518] 0.7978736601946161 2020-01-23 10:35:54,123 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 10:35:57,021 Loss: 0.163, MeanIU: 0.7979, Best_mIoU: 0.8033 2020-01-23 10:35:57,022 [0.98436086 0.87094216 0.93478097 0.58798563 0.64561136 0.70716943 0.75270548 0.82637526 0.93108823 0.64316581 0.95323301 0.84296612 0.63817268 0.95687612 0.76192188 0.87704648 0.77750263 0.67032026 0.79737518] 2020-01-23 10:35:58,913 Epoch: [477/484] Iter:[0/247], Time: 1.88, lr: [0.00022091612179404727], Loss: 0.096867 2020-01-23 10:36:07,013 Epoch: [477/484] Iter:[10/247], Time: 0.91, lr: [0.00021976584921518244], Loss: 0.116482 2020-01-23 10:36:15,409 Epoch: [477/484] Iter:[20/247], Time: 0.88, lr: [0.00021861490728551276], Loss: 0.107135 2020-01-23 10:36:23,538 Epoch: [477/484] Iter:[30/247], Time: 0.86, lr: [0.00021746329169544215], Loss: 0.109641 2020-01-23 10:36:31,599 Epoch: [477/484] Iter:[40/247], Time: 0.84, lr: [0.00021631099808209022], Loss: 0.108900 2020-01-23 10:36:39,642 Epoch: [477/484] Iter:[50/247], Time: 0.84, lr: [0.000215158022028319], Loss: 0.111877 2020-01-23 10:36:47,891 Epoch: [477/484] Iter:[60/247], Time: 0.83, lr: [0.00021400435906171965], Loss: 0.114716 2020-01-23 10:36:56,021 Epoch: [477/484] Iter:[70/247], Time: 0.83, lr: [0.0002128500046535983], Loss: 0.115094 2020-01-23 10:37:04,138 Epoch: [477/484] Iter:[80/247], Time: 0.83, lr: [0.00021169495421791116], Loss: 0.114057 2020-01-23 10:37:12,314 Epoch: [477/484] Iter:[90/247], Time: 0.83, lr: [0.00021053920311019873], Loss: 0.115403 2020-01-23 10:37:20,649 Epoch: [477/484] Iter:[100/247], Time: 0.83, lr: [0.00020938274662646804], Loss: 0.117711 2020-01-23 10:37:28,768 Epoch: [477/484] Iter:[110/247], Time: 0.83, lr: [0.0002082255800020718], Loss: 0.118469 2020-01-23 10:37:36,907 Epoch: [477/484] Iter:[120/247], Time: 0.83, lr: [0.00020706769841053417], Loss: 0.117450 2020-01-23 10:37:44,886 Epoch: [477/484] Iter:[130/247], Time: 0.82, lr: [0.0002059090969623716], Loss: 0.119053 2020-01-23 10:37:52,941 Epoch: [477/484] Iter:[140/247], Time: 0.82, lr: [0.00020474977070385837], Loss: 0.120153 2020-01-23 10:38:01,111 Epoch: [477/484] Iter:[150/247], Time: 0.82, lr: [0.00020358971461578544], Loss: 0.120780 2020-01-23 10:38:09,493 Epoch: [477/484] Iter:[160/247], Time: 0.82, lr: [0.00020242892361216173], Loss: 0.120566 2020-01-23 10:38:17,653 Epoch: [477/484] Iter:[170/247], Time: 0.82, lr: [0.0002012673925389069], Loss: 0.121247 2020-01-23 10:38:25,817 Epoch: [477/484] Iter:[180/247], Time: 0.82, lr: [0.00020010511617248587], Loss: 0.121379 2020-01-23 10:38:33,962 Epoch: [477/484] Iter:[190/247], Time: 0.82, lr: [0.00019894208921852313], Loss: 0.121150 2020-01-23 10:38:42,212 Epoch: [477/484] Iter:[200/247], Time: 0.82, lr: [0.00019777830631037988], Loss: 0.121696 2020-01-23 10:38:50,437 Epoch: [477/484] Iter:[210/247], Time: 0.82, lr: [0.00019661376200767755], Loss: 0.120519 2020-01-23 10:38:58,621 Epoch: [477/484] Iter:[220/247], Time: 0.82, lr: [0.00019544845079480484], Loss: 0.120574 2020-01-23 10:39:06,905 Epoch: [477/484] Iter:[230/247], Time: 0.82, lr: [0.00019428236707935906], Loss: 0.119238 2020-01-23 10:39:14,983 Epoch: [477/484] Iter:[240/247], Time: 0.82, lr: [0.00019311550519057017], Loss: 0.119053 2020-01-23 10:42:37,601 0 [0.98398294 0.86679832 0.93406855 0.56691882 0.6333281 0.70195324 0.75093721 0.82222477 0.93030906 0.63697811 0.9528853 0.84030586 0.63557602 0.95357125 0.70890745 0.83924365 0.63282867 0.67102656 0.79624532] 0.7820046944766762 2020-01-23 10:42:37,601 1 [0.98418876 0.86869725 0.93450913 0.57023798 0.63873659 0.7060927 0.7535753 0.8255608 0.93104661 0.64131875 0.95372515 0.84345351 0.64368614 0.95793815 0.77442573 0.84302023 0.65642473 0.67694821 0.79869442] 0.7895936916075983 2020-01-23 10:42:37,602 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 10:42:40,430 Loss: 0.165, MeanIU: 0.7896, Best_mIoU: 0.8033 2020-01-23 10:42:40,430 [0.98418876 0.86869725 0.93450913 0.57023798 0.63873659 0.7060927 0.7535753 0.8255608 0.93104661 0.64131875 0.95372515 0.84345351 0.64368614 0.95793815 0.77442573 0.84302023 0.65642473 0.67694821 0.79869442] 2020-01-23 10:42:42,239 Epoch: [478/484] Iter:[0/247], Time: 1.80, lr: [0.00019229823577899684], Loss: 0.129491 2020-01-23 10:42:50,291 Epoch: [478/484] Iter:[10/247], Time: 0.90, lr: [0.000191130037753811], Loss: 0.101415 2020-01-23 10:42:58,407 Epoch: [478/484] Iter:[20/247], Time: 0.86, lr: [0.00018996104583979007], Loss: 0.113695 2020-01-23 10:43:06,691 Epoch: [478/484] Iter:[30/247], Time: 0.85, lr: [0.00018879125406162919], Loss: 0.110362 2020-01-23 10:43:14,840 Epoch: [478/484] Iter:[40/247], Time: 0.84, lr: [0.00018762065635756573], Loss: 0.114077 2020-01-23 10:43:23,121 Epoch: [478/484] Iter:[50/247], Time: 0.84, lr: [0.00018644924657753258], Loss: 0.113466 2020-01-23 10:43:31,201 Epoch: [478/484] Iter:[60/247], Time: 0.83, lr: [0.0001852770184812348], Loss: 0.116276 2020-01-23 10:43:39,181 Epoch: [478/484] Iter:[70/247], Time: 0.83, lr: [0.00018410396573618638], Loss: 0.117617 2020-01-23 10:43:47,199 Epoch: [478/484] Iter:[80/247], Time: 0.82, lr: [0.00018293008191569052], Loss: 0.117668 2020-01-23 10:43:55,242 Epoch: [478/484] Iter:[90/247], Time: 0.82, lr: [0.0001817553604967444], Loss: 0.117995 2020-01-23 10:44:03,382 Epoch: [478/484] Iter:[100/247], Time: 0.82, lr: [0.00018057979485790624], Loss: 0.118279 2020-01-23 10:44:11,639 Epoch: [478/484] Iter:[110/247], Time: 0.82, lr: [0.0001794033782770738], Loss: 0.118827 2020-01-23 10:44:19,766 Epoch: [478/484] Iter:[120/247], Time: 0.82, lr: [0.00017822610392922154], Loss: 0.118753 2020-01-23 10:44:27,854 Epoch: [478/484] Iter:[130/247], Time: 0.82, lr: [0.00017704796488404463], Loss: 0.118354 2020-01-23 10:44:36,145 Epoch: [478/484] Iter:[140/247], Time: 0.82, lr: [0.00017586895410355658], Loss: 0.119227 2020-01-23 10:44:44,284 Epoch: [478/484] Iter:[150/247], Time: 0.82, lr: [0.00017468906443958806], Loss: 0.118609 2020-01-23 10:44:52,303 Epoch: [478/484] Iter:[160/247], Time: 0.82, lr: [0.00017350828863123437], Loss: 0.118455 2020-01-23 10:45:00,349 Epoch: [478/484] Iter:[170/247], Time: 0.82, lr: [0.0001723266193021979], Loss: 0.117734 2020-01-23 10:45:08,444 Epoch: [478/484] Iter:[180/247], Time: 0.82, lr: [0.00017114404895807338], Loss: 0.118801 2020-01-23 10:45:16,533 Epoch: [478/484] Iter:[190/247], Time: 0.82, lr: [0.00016996056998352168], Loss: 0.119049 2020-01-23 10:45:24,772 Epoch: [478/484] Iter:[200/247], Time: 0.82, lr: [0.00016877617463937977], Loss: 0.118854 2020-01-23 10:45:32,734 Epoch: [478/484] Iter:[210/247], Time: 0.82, lr: [0.00016759085505965412], Loss: 0.118415 2020-01-23 10:45:40,914 Epoch: [478/484] Iter:[220/247], Time: 0.82, lr: [0.00016640460324843313], Loss: 0.118679 2020-01-23 10:45:48,956 Epoch: [478/484] Iter:[230/247], Time: 0.82, lr: [0.00016521741107669908], Loss: 0.118216 2020-01-23 10:45:57,147 Epoch: [478/484] Iter:[240/247], Time: 0.82, lr: [0.00016402927027901992], Loss: 0.117719 2020-01-23 10:49:34,408 0 [0.9841543 0.86796994 0.93418433 0.57255827 0.63485811 0.70366195 0.74884056 0.82493559 0.93030724 0.64167735 0.95297546 0.84228228 0.6416854 0.95351417 0.70785445 0.85559025 0.68958127 0.66382912 0.79403551] 0.7865523973924439 2020-01-23 10:49:34,408 1 [0.98431336 0.86979241 0.93476466 0.57784828 0.6399756 0.70746188 0.75174547 0.82828127 0.9310032 0.64387285 0.95382535 0.84434411 0.64735739 0.95690231 0.75529809 0.87356446 0.76344711 0.67239885 0.79580435] 0.7964211054269583 2020-01-23 10:49:34,409 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 10:49:37,240 Loss: 0.164, MeanIU: 0.7964, Best_mIoU: 0.8033 2020-01-23 10:49:37,241 [0.98431336 0.86979241 0.93476466 0.57784828 0.6399756 0.70746188 0.75174547 0.82828127 0.9310032 0.64387285 0.95382535 0.84434411 0.64735739 0.95690231 0.75529809 0.87356446 0.76344711 0.67239885 0.79580435] 2020-01-23 10:49:39,217 Epoch: [479/484] Iter:[0/247], Time: 1.97, lr: [0.0001631970027929206], Loss: 0.081131 2020-01-23 10:49:47,378 Epoch: [479/484] Iter:[10/247], Time: 0.92, lr: [0.0001620072299644936], Loss: 0.109495 2020-01-23 10:49:55,426 Epoch: [479/484] Iter:[20/247], Time: 0.87, lr: [0.0001608164854850733], Loss: 0.117302 2020-01-23 10:50:03,545 Epoch: [479/484] Iter:[30/247], Time: 0.85, lr: [0.00015962476055399712], Loss: 0.125331 2020-01-23 10:50:11,566 Epoch: [479/484] Iter:[40/247], Time: 0.84, lr: [0.0001584320462171676], Loss: 0.122308 2020-01-23 10:50:19,906 Epoch: [479/484] Iter:[50/247], Time: 0.84, lr: [0.00015723833336305716], Loss: 0.124601 2020-01-23 10:50:27,995 Epoch: [479/484] Iter:[60/247], Time: 0.83, lr: [0.00015604361271860095], Loss: 0.117816 2020-01-23 10:50:36,220 Epoch: [479/484] Iter:[70/247], Time: 0.83, lr: [0.00015484787484492142], Loss: 0.118768 2020-01-23 10:50:44,292 Epoch: [479/484] Iter:[80/247], Time: 0.83, lr: [0.00015365111013291904], Loss: 0.117914 2020-01-23 10:50:52,559 Epoch: [479/484] Iter:[90/247], Time: 0.83, lr: [0.00015245330879870613], Loss: 0.118958 2020-01-23 10:51:00,698 Epoch: [479/484] Iter:[100/247], Time: 0.83, lr: [0.00015125446087886106], Loss: 0.119236 2020-01-23 10:51:08,767 Epoch: [479/484] Iter:[110/247], Time: 0.82, lr: [0.00015005455622553516], Loss: 0.118254 2020-01-23 10:51:16,997 Epoch: [479/484] Iter:[120/247], Time: 0.82, lr: [0.00014885358450135528], Loss: 0.118614 2020-01-23 10:51:25,038 Epoch: [479/484] Iter:[130/247], Time: 0.82, lr: [0.00014765153517416488], Loss: 0.117408 2020-01-23 10:51:33,261 Epoch: [479/484] Iter:[140/247], Time: 0.82, lr: [0.00014644839751154373], Loss: 0.117487 2020-01-23 10:51:41,368 Epoch: [479/484] Iter:[150/247], Time: 0.82, lr: [0.00014524416057514833], Loss: 0.118239 2020-01-23 10:51:49,490 Epoch: [479/484] Iter:[160/247], Time: 0.82, lr: [0.00014403881321481263], Loss: 0.118526 2020-01-23 10:51:57,698 Epoch: [479/484] Iter:[170/247], Time: 0.82, lr: [0.00014283234406244923], Loss: 0.119753 2020-01-23 10:52:05,795 Epoch: [479/484] Iter:[180/247], Time: 0.82, lr: [0.00014162474152568992], Loss: 0.120708 2020-01-23 10:52:14,097 Epoch: [479/484] Iter:[190/247], Time: 0.82, lr: [0.00014041599378130485], Loss: 0.122078 2020-01-23 10:52:22,266 Epoch: [479/484] Iter:[200/247], Time: 0.82, lr: [0.00013920608876833736], Loss: 0.122148 2020-01-23 10:52:30,324 Epoch: [479/484] Iter:[210/247], Time: 0.82, lr: [0.00013799501418099294], Loss: 0.121960 2020-01-23 10:52:38,570 Epoch: [479/484] Iter:[220/247], Time: 0.82, lr: [0.00013678275746121856], Loss: 0.121769 2020-01-23 10:52:46,717 Epoch: [479/484] Iter:[230/247], Time: 0.82, lr: [0.00013556930579099928], Loss: 0.122217 2020-01-23 10:52:54,774 Epoch: [479/484] Iter:[240/247], Time: 0.82, lr: [0.0001343546460843396], Loss: 0.122411 2020-01-23 10:56:39,863 0 [0.98430026 0.86900755 0.93408138 0.57856017 0.63171872 0.702891 0.74812163 0.82554758 0.93012189 0.63710888 0.95304822 0.84224078 0.64088722 0.95347816 0.7051187 0.85940805 0.69308415 0.67099572 0.79580662] 0.7871329829010906 2020-01-23 10:56:39,863 1 [0.98445291 0.87029548 0.93464216 0.58061946 0.63857153 0.70688833 0.75077554 0.82887789 0.93085551 0.63905946 0.95367792 0.84462599 0.64739221 0.95614767 0.74526927 0.87664288 0.75709174 0.67864983 0.79785673] 0.7959153960351941 2020-01-23 10:56:39,864 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 10:56:42,710 Loss: 0.164, MeanIU: 0.7959, Best_mIoU: 0.8033 2020-01-23 10:56:42,710 [0.98445291 0.87029548 0.93464216 0.58061946 0.63857153 0.70688833 0.75077554 0.82887789 0.93085551 0.63905946 0.95367792 0.84462599 0.64739221 0.95614767 0.74526927 0.87664288 0.75709174 0.67864983 0.79785673] 2020-01-23 10:56:44,580 Epoch: [480/484] Iter:[0/247], Time: 1.86, lr: [0.00013350365836381595], Loss: 0.107470 2020-01-23 10:56:52,588 Epoch: [480/484] Iter:[10/247], Time: 0.90, lr: [0.00013228691417635965], Loss: 0.116083 2020-01-23 10:57:00,732 Epoch: [480/484] Iter:[20/247], Time: 0.86, lr: [0.0001310689252151365], Loss: 0.126424 2020-01-23 10:57:08,868 Epoch: [480/484] Iter:[30/247], Time: 0.84, lr: [0.00012984967732718321], Loss: 0.118923 2020-01-23 10:57:16,895 Epoch: [480/484] Iter:[40/247], Time: 0.83, lr: [0.00012862915604910919], Loss: 0.116873 2020-01-23 10:57:25,094 Epoch: [480/484] Iter:[50/247], Time: 0.83, lr: [0.00012740734659695226], Loss: 0.113165 2020-01-23 10:57:33,221 Epoch: [480/484] Iter:[60/247], Time: 0.83, lr: [0.00012618423385556477], Loss: 0.113899 2020-01-23 10:57:41,323 Epoch: [480/484] Iter:[70/247], Time: 0.83, lr: [0.00012495980236755655], Loss: 0.113630 2020-01-23 10:57:49,583 Epoch: [480/484] Iter:[80/247], Time: 0.83, lr: [0.00012373403632171887], Loss: 0.116201 2020-01-23 10:57:57,716 Epoch: [480/484] Iter:[90/247], Time: 0.82, lr: [0.00012250691954094234], Loss: 0.114807 2020-01-23 10:58:05,739 Epoch: [480/484] Iter:[100/247], Time: 0.82, lr: [0.00012127843546958239], Loss: 0.116589 2020-01-23 10:58:13,884 Epoch: [480/484] Iter:[110/247], Time: 0.82, lr: [0.00012004856716022444], Loss: 0.116911 2020-01-23 10:58:22,008 Epoch: [480/484] Iter:[120/247], Time: 0.82, lr: [0.0001188172972598559], Loss: 0.115955 2020-01-23 10:58:29,983 Epoch: [480/484] Iter:[130/247], Time: 0.82, lr: [0.0001175846079953575], Loss: 0.118248 2020-01-23 10:58:38,051 Epoch: [480/484] Iter:[140/247], Time: 0.82, lr: [0.00011635048115832787], Loss: 0.117357 2020-01-23 10:58:46,236 Epoch: [480/484] Iter:[150/247], Time: 0.82, lr: [0.00011511489808914687], Loss: 0.116876 2020-01-23 10:58:54,361 Epoch: [480/484] Iter:[160/247], Time: 0.82, lr: [0.00011387783966028567], Loss: 0.116245 2020-01-23 10:59:02,667 Epoch: [480/484] Iter:[170/247], Time: 0.82, lr: [0.00011263928625876283], Loss: 0.114778 2020-01-23 10:59:10,991 Epoch: [480/484] Iter:[180/247], Time: 0.82, lr: [0.00011139921776774728], Loss: 0.116359 2020-01-23 10:59:19,340 Epoch: [480/484] Iter:[190/247], Time: 0.82, lr: [0.00011015761354720029], Loss: 0.116121 2020-01-23 10:59:27,528 Epoch: [480/484] Iter:[200/247], Time: 0.82, lr: [0.00010891445241354888], Loss: 0.116680 2020-01-23 10:59:35,698 Epoch: [480/484] Iter:[210/247], Time: 0.82, lr: [0.00010766971261827402], Loss: 0.118730 2020-01-23 10:59:43,748 Epoch: [480/484] Iter:[220/247], Time: 0.82, lr: [0.00010642337182539646], Loss: 0.119317 2020-01-23 10:59:51,778 Epoch: [480/484] Iter:[230/247], Time: 0.82, lr: [0.00010517540708773274], Loss: 0.119150 2020-01-23 10:59:59,758 Epoch: [480/484] Iter:[240/247], Time: 0.82, lr: [0.00010392579482189281], Loss: 0.119178 2020-01-23 11:03:36,228 0 [0.98445085 0.86983538 0.93409369 0.57275511 0.63350737 0.70436549 0.74866048 0.82440248 0.93007521 0.64050603 0.95343372 0.84184206 0.64134026 0.95347463 0.70970861 0.86232047 0.70268417 0.66376538 0.79512108] 0.7877022348042918 2020-01-23 11:03:36,229 1 [0.98460798 0.87151552 0.93466816 0.57819068 0.64017744 0.70858521 0.75095473 0.82756112 0.93086834 0.6440191 0.95434979 0.84409075 0.64688333 0.95656219 0.7580944 0.88014135 0.766376 0.67053394 0.79709075] 0.7971195150766698 2020-01-23 11:03:36,229 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 11:03:39,151 Loss: 0.164, MeanIU: 0.7971, Best_mIoU: 0.8033 2020-01-23 11:03:39,151 [0.98460798 0.87151552 0.93466816 0.57819068 0.64017744 0.70858521 0.75095473 0.82756112 0.93086834 0.6440191 0.95434979 0.84409075 0.64688333 0.95656219 0.7580944 0.88014135 0.766376 0.67053394 0.79709075] 2020-01-23 11:03:41,071 Epoch: [481/484] Iter:[0/247], Time: 1.91, lr: [0.00010305007300023879], Loss: 0.059279 2020-01-23 11:03:49,454 Epoch: [481/484] Iter:[10/247], Time: 0.94, lr: [0.00010179760392336366], Loss: 0.119549 2020-01-23 11:03:57,643 Epoch: [481/484] Iter:[20/247], Time: 0.88, lr: [0.00010054342025384094], Loss: 0.124599 2020-01-23 11:04:05,816 Epoch: [481/484] Iter:[30/247], Time: 0.86, lr: [9.928749581301605e-05], Loss: 0.122993 2020-01-23 11:04:14,003 Epoch: [481/484] Iter:[40/247], Time: 0.85, lr: [9.802980364841368e-05], Loss: 0.120393 2020-01-23 11:04:22,418 Epoch: [481/484] Iter:[50/247], Time: 0.85, lr: [9.677031599947501e-05], Loss: 0.121214 2020-01-23 11:04:30,550 Epoch: [481/484] Iter:[60/247], Time: 0.84, lr: [9.55090042612875e-05], Loss: 0.121931 2020-01-23 11:04:38,967 Epoch: [481/484] Iter:[70/247], Time: 0.84, lr: [9.424583894610179e-05], Loss: 0.119543 2020-01-23 11:04:47,041 Epoch: [481/484] Iter:[80/247], Time: 0.84, lr: [9.29807896425223e-05], Loss: 0.122222 2020-01-23 11:04:55,304 Epoch: [481/484] Iter:[90/247], Time: 0.84, lr: [9.171382497213674e-05], Loss: 0.121666 2020-01-23 11:05:03,461 Epoch: [481/484] Iter:[100/247], Time: 0.83, lr: [9.044491254343908e-05], Loss: 0.123080 2020-01-23 11:05:11,649 Epoch: [481/484] Iter:[110/247], Time: 0.83, lr: [8.917401890277648e-05], Loss: 0.124254 2020-01-23 11:05:19,796 Epoch: [481/484] Iter:[120/247], Time: 0.83, lr: [8.790110948212332e-05], Loss: 0.125276 2020-01-23 11:05:27,849 Epoch: [481/484] Iter:[130/247], Time: 0.83, lr: [8.662614854340343e-05], Loss: 0.125216 2020-01-23 11:05:36,058 Epoch: [481/484] Iter:[140/247], Time: 0.83, lr: [8.534909911905492e-05], Loss: 0.124686 2020-01-23 11:05:44,159 Epoch: [481/484] Iter:[150/247], Time: 0.83, lr: [8.406992294856246e-05], Loss: 0.124287 2020-01-23 11:05:52,191 Epoch: [481/484] Iter:[160/247], Time: 0.83, lr: [8.278858041055368e-05], Loss: 0.123899 2020-01-23 11:06:00,210 Epoch: [481/484] Iter:[170/247], Time: 0.82, lr: [8.150503045012669e-05], Loss: 0.123689 2020-01-23 11:06:08,479 Epoch: [481/484] Iter:[180/247], Time: 0.82, lr: [8.021923050092539e-05], Loss: 0.123981 2020-01-23 11:06:16,615 Epoch: [481/484] Iter:[190/247], Time: 0.82, lr: [7.893113640154154e-05], Loss: 0.122929 2020-01-23 11:06:24,692 Epoch: [481/484] Iter:[200/247], Time: 0.82, lr: [7.764070230566066e-05], Loss: 0.123816 2020-01-23 11:06:32,810 Epoch: [481/484] Iter:[210/247], Time: 0.82, lr: [7.634788058541861e-05], Loss: 0.123057 2020-01-23 11:06:40,988 Epoch: [481/484] Iter:[220/247], Time: 0.82, lr: [7.505262172725853e-05], Loss: 0.123191 2020-01-23 11:06:49,084 Epoch: [481/484] Iter:[230/247], Time: 0.82, lr: [7.375487421961049e-05], Loss: 0.123094 2020-01-23 11:06:57,169 Epoch: [481/484] Iter:[240/247], Time: 0.82, lr: [7.245458443151907e-05], Loss: 0.122341 2020-01-23 11:10:30,581 0 [0.98422507 0.86937829 0.93397827 0.57422895 0.63350519 0.70429236 0.74893811 0.82422439 0.92985008 0.64135694 0.95303837 0.84191301 0.64024099 0.95365719 0.71253503 0.85804545 0.69231959 0.66186598 0.79317115] 0.7868823371116404 2020-01-23 11:10:30,582 1 [0.98445409 0.87110562 0.93468324 0.57607456 0.63886157 0.70849035 0.75044198 0.82732651 0.93071402 0.64468807 0.95383681 0.84403096 0.64442819 0.95694307 0.76196315 0.87660772 0.76019372 0.67061543 0.79534337] 0.796358022659045 2020-01-23 11:10:30,582 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 11:10:33,413 Loss: 0.164, MeanIU: 0.7964, Best_mIoU: 0.8033 2020-01-23 11:10:33,414 [0.98445409 0.87110562 0.93468324 0.57607456 0.63886157 0.70849035 0.75044198 0.82732651 0.93071402 0.64468807 0.95383681 0.84403096 0.64442819 0.95694307 0.76196315 0.87660772 0.76019372 0.67061543 0.79534337] 2020-01-23 11:10:35,334 Epoch: [482/484] Iter:[0/247], Time: 1.91, lr: [7.154283909292469e-05], Loss: 0.166780 2020-01-23 11:10:43,746 Epoch: [482/484] Iter:[10/247], Time: 0.94, lr: [7.023809790436952e-05], Loss: 0.114268 2020-01-23 11:10:51,784 Epoch: [482/484] Iter:[20/247], Time: 0.87, lr: [6.893065798093745e-05], Loss: 0.119852 2020-01-23 11:11:00,054 Epoch: [482/484] Iter:[30/247], Time: 0.86, lr: [6.762045661843488e-05], Loss: 0.107438 2020-01-23 11:11:08,691 Epoch: [482/484] Iter:[40/247], Time: 0.86, lr: [6.630742827105917e-05], Loss: 0.109519 2020-01-23 11:11:17,090 Epoch: [482/484] Iter:[50/247], Time: 0.86, lr: [6.499150435710936e-05], Loss: 0.111015 2020-01-23 11:11:25,394 Epoch: [482/484] Iter:[60/247], Time: 0.85, lr: [6.367261304669901e-05], Loss: 0.109332 2020-01-23 11:11:33,565 Epoch: [482/484] Iter:[70/247], Time: 0.85, lr: [6.23506790294124e-05], Loss: 0.109176 2020-01-23 11:11:41,787 Epoch: [482/484] Iter:[80/247], Time: 0.84, lr: [6.102562325942771e-05], Loss: 0.110645 2020-01-23 11:11:49,907 Epoch: [482/484] Iter:[90/247], Time: 0.84, lr: [5.969736267538652e-05], Loss: 0.110461 2020-01-23 11:11:58,131 Epoch: [482/484] Iter:[100/247], Time: 0.84, lr: [5.836580989175592e-05], Loss: 0.112154 2020-01-23 11:12:06,324 Epoch: [482/484] Iter:[110/247], Time: 0.84, lr: [5.703087285804628e-05], Loss: 0.111316 2020-01-23 11:12:14,492 Epoch: [482/484] Iter:[120/247], Time: 0.84, lr: [5.5692454481549096e-05], Loss: 0.112158 2020-01-23 11:12:22,475 Epoch: [482/484] Iter:[130/247], Time: 0.83, lr: [5.435045220865801e-05], Loss: 0.112527 2020-01-23 11:12:30,500 Epoch: [482/484] Iter:[140/247], Time: 0.83, lr: [5.300475755893817e-05], Loss: 0.113887 2020-01-23 11:12:38,622 Epoch: [482/484] Iter:[150/247], Time: 0.83, lr: [5.165525560510407e-05], Loss: 0.112811 2020-01-23 11:12:46,801 Epoch: [482/484] Iter:[160/247], Time: 0.83, lr: [5.0301824390915753e-05], Loss: 0.115787 2020-01-23 11:12:55,062 Epoch: [482/484] Iter:[170/247], Time: 0.83, lr: [4.8944334277432656e-05], Loss: 0.116089 2020-01-23 11:13:03,269 Epoch: [482/484] Iter:[180/247], Time: 0.83, lr: [4.7582647206382684e-05], Loss: 0.116651 2020-01-23 11:13:11,431 Epoch: [482/484] Iter:[190/247], Time: 0.83, lr: [4.6216615867088406e-05], Loss: 0.116704 2020-01-23 11:13:19,684 Epoch: [482/484] Iter:[200/247], Time: 0.83, lr: [4.4846082750790883e-05], Loss: 0.117078 2020-01-23 11:13:27,943 Epoch: [482/484] Iter:[210/247], Time: 0.83, lr: [4.3470879072721804e-05], Loss: 0.117720 2020-01-23 11:13:36,182 Epoch: [482/484] Iter:[220/247], Time: 0.83, lr: [4.209082353816928e-05], Loss: 0.118627 2020-01-23 11:13:44,359 Epoch: [482/484] Iter:[230/247], Time: 0.83, lr: [4.0705720923348604e-05], Loss: 0.118686 2020-01-23 11:13:52,441 Epoch: [482/484] Iter:[240/247], Time: 0.83, lr: [3.931536043525206e-05], Loss: 0.119942 2020-01-23 11:17:11,007 0 [0.98414966 0.86895431 0.93426231 0.57670458 0.634724 0.70516927 0.75071086 0.8250581 0.93018378 0.64025838 0.95302315 0.84245084 0.64344523 0.95362048 0.71624589 0.85770536 0.6860041 0.66650235 0.79521642] 0.7875994250409313 2020-01-23 11:17:11,008 1 [0.9843783 0.87077118 0.93492977 0.58157962 0.64100609 0.70981016 0.75212876 0.82800797 0.93097216 0.64313357 0.95364184 0.84508425 0.64973645 0.95708874 0.76677151 0.8753902 0.75322363 0.67198745 0.79760631] 0.7972235767655834 2020-01-23 11:17:11,008 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 11:17:13,936 Loss: 0.163, MeanIU: 0.7972, Best_mIoU: 0.8033 2020-01-23 11:17:13,936 [0.9843783 0.87077118 0.93492977 0.58157962 0.64100609 0.70981016 0.75212876 0.82800797 0.93097216 0.64313357 0.95364184 0.84508425 0.64973645 0.95708874 0.76677151 0.8753902 0.75322363 0.67198745 0.79760631] 2020-01-23 11:17:15,794 Epoch: [483/484] Iter:[0/247], Time: 1.85, lr: [3.833885818715131e-05], Loss: 0.192975 2020-01-23 11:17:23,939 Epoch: [483/484] Iter:[10/247], Time: 0.91, lr: [3.693902498956716e-05], Loss: 0.139059 2020-01-23 11:17:32,061 Epoch: [483/484] Iter:[20/247], Time: 0.86, lr: [3.5533270994975924e-05], Loss: 0.125767 2020-01-23 11:17:40,075 Epoch: [483/484] Iter:[30/247], Time: 0.84, lr: [3.4121308476897076e-05], Loss: 0.126252 2020-01-23 11:17:48,402 Epoch: [483/484] Iter:[40/247], Time: 0.84, lr: [3.270282177209703e-05], Loss: 0.120885 2020-01-23 11:17:56,485 Epoch: [483/484] Iter:[50/247], Time: 0.83, lr: [3.127746308071114e-05], Loss: 0.113956 2020-01-23 11:18:04,567 Epoch: [483/484] Iter:[60/247], Time: 0.83, lr: [2.9844847388417388e-05], Loss: 0.111658 2020-01-23 11:18:12,687 Epoch: [483/484] Iter:[70/247], Time: 0.83, lr: [2.8404546269919003e-05], Loss: 0.113975 2020-01-23 11:18:20,838 Epoch: [483/484] Iter:[80/247], Time: 0.83, lr: [2.695608024967217e-05], Loss: 0.114058 2020-01-23 11:18:28,990 Epoch: [483/484] Iter:[90/247], Time: 0.82, lr: [2.5498909276512854e-05], Loss: 0.114804 2020-01-23 11:18:37,167 Epoch: [483/484] Iter:[100/247], Time: 0.82, lr: [2.403242069550214e-05], Loss: 0.115689 2020-01-23 11:18:45,566 Epoch: [483/484] Iter:[110/247], Time: 0.83, lr: [2.2555913842252914e-05], Loss: 0.117390 2020-01-23 11:18:53,909 Epoch: [483/484] Iter:[120/247], Time: 0.83, lr: [2.106857999227624e-05], Loss: 0.118007 2020-01-23 11:19:02,083 Epoch: [483/484] Iter:[130/247], Time: 0.83, lr: [1.9569475783004125e-05], Loss: 0.116486 2020-01-23 11:19:10,185 Epoch: [483/484] Iter:[140/247], Time: 0.82, lr: [1.8057487234054166e-05], Loss: 0.116207 2020-01-23 11:19:18,276 Epoch: [483/484] Iter:[150/247], Time: 0.82, lr: [1.653127983166872e-05], Loss: 0.116788 2020-01-23 11:19:26,618 Epoch: [483/484] Iter:[160/247], Time: 0.82, lr: [1.498922725005196e-05], Loss: 0.117709 2020-01-23 11:19:34,735 Epoch: [483/484] Iter:[170/247], Time: 0.82, lr: [1.3429305983725939e-05], Loss: 0.117533 2020-01-23 11:19:42,898 Epoch: [483/484] Iter:[180/247], Time: 0.82, lr: [1.1848932870667405e-05], Loss: 0.117266 2020-01-23 11:19:51,219 Epoch: [483/484] Iter:[190/247], Time: 0.82, lr: [1.0244700976647234e-05], Loss: 0.117899 2020-01-23 11:19:59,265 Epoch: [483/484] Iter:[200/247], Time: 0.82, lr: [8.611920037976371e-06], Loss: 0.117801 2020-01-23 11:20:07,402 Epoch: [483/484] Iter:[210/247], Time: 0.82, lr: [6.943740278341014e-06], Loss: 0.117981 2020-01-23 11:20:15,736 Epoch: [483/484] Iter:[220/247], Time: 0.82, lr: [5.229248786907857e-06], Loss: 0.119557 2020-01-23 11:20:23,976 Epoch: [483/484] Iter:[230/247], Time: 0.82, lr: [3.4483865763123378e-06], Loss: 0.119375 2020-01-23 11:20:32,168 Epoch: [483/484] Iter:[240/247], Time: 0.82, lr: [1.55167281912151e-06], Loss: 0.119287 2020-01-23 11:24:08,399 0 [0.98427684 0.86914601 0.93384861 0.57542279 0.63306998 0.70514902 0.75020965 0.82584636 0.93002661 0.63876686 0.9526544 0.84269889 0.64094631 0.95342941 0.71083969 0.8606712 0.7045397 0.6648826 0.79501391] 0.7879704663729828 2020-01-23 11:24:08,400 1 [0.98446208 0.87073235 0.93450846 0.57955524 0.63832697 0.70909732 0.75140754 0.82827462 0.93078488 0.64213624 0.95328786 0.84514085 0.64734001 0.95666015 0.75642446 0.87660393 0.77795041 0.66563148 0.79636404] 0.7970888892500124 2020-01-23 11:24:08,401 => saving checkpoint to output/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484checkpoint.pth.tar 2020-01-23 11:24:11,270 Loss: 0.165, MeanIU: 0.7971, Best_mIoU: 0.8033 2020-01-23 11:24:11,270 [0.98446208 0.87073235 0.93450846 0.57955524 0.63832697 0.70909732 0.75140754 0.82827462 0.93078488 0.64213624 0.95328786 0.84514085 0.64734001 0.95666015 0.75642446 0.87660393 0.77795041 0.66563148 0.79636404] 2020-01-23 11:24:11,718 Hours: 2 2020-01-23 11:24:11,718 Done

Jan 24 '20 04:01 verymadmatt

@verymadmatt We perform class-balance for all the experiments on Cityscapes. We only use the Cityscapes train set for training. First, could you provide more details about your environmental information? We expect you to conduct experiments with Pytorch1.1. Second, please ensure that you have reproduced the performance of the HRNet baseline.

Jan 26 '20 14:01 PkuRainBow

@PkuRainBow Thanks for your reply. I'm using python 3.6, pytorch 1.1 and 4 P100 GPUs. Others just followed the 'requirements.txt' file. The mIoU increased to 0.8064 when i turned class_balance on, but still 1% lower than the reported. I will try to reproduce HRNet first. But it will take me ~3 days. Would the performance be of any difference if I double the BS and base learning rate accordingly? Any advice will be much appreciated.

Jan 28 '20 08:01 verymadmatt

@hsfzxjy Please check the possible reasons.

Jan 28 '20 15:01 PkuRainBow

@hsfzxjy @PkuRainBow FYI. I have reproduced the performance of HRNetV2 with the provided config file. The mIoU reached 0.8086 as reported (0.809). I tried HRNetV2+OCR with random_brightness turned on as the paper suggested. But the mIoU downgraded from 0.8064 to 0.8033. Not sure if I missed anything for HRNetV2+OCR, any advice will be much appreciated.

Jan 30 '20 13:01 verymadmatt

@verymadmatt Please be patient. We will check the possible problems and reply to you latter.

Jan 31 '20 04:01 PkuRainBow

@PkuRainBow Thanks for your reply. I'm using python 3.6, pytorch 1.1 and 4 P100 GPUs. Others just followed the 'requirements.txt' file. The mIoU increased to 0.8064 when i turned class_balance on, but still 1% lower than the reported. I will try to reproduce HRNet first. But it will take me ~3 days. Would the performance be of any difference if I double the BS and base learning rate accordingly? Any advice will be much appreciated.

hello, which version of cuda are you using?

Feb 06 '20 02:02 laojiangwei

@PkuRainBow Thanks for your reply. I'm using python 3.6, pytorch 1.1 and 4 P100 GPUs. Others just followed the 'requirements.txt' file. The mIoU increased to 0.8064 when i turned class_balance on, but still 1% lower than the reported. I will try to reproduce HRNet first. But it will take me ~3 days. Would the performance be of any difference if I double the BS and base learning rate accordingly? Any advice will be much appreciated.

hello, which version of cuda are you using?

8.0. Which version should I use?

Feb 06 '20 04:02 verymadmatt

@PkuRainBow Thanks for your reply. I'm using python 3.6, pytorch 1.1 and 4 P100 GPUs. Others just followed the 'requirements.txt' file. The mIoU increased to 0.8064 when i turned class_balance on, but still 1% lower than the reported. I will try to reproduce HRNet first. But it will take me ~3 days. Would the performance be of any difference if I double the BS and base learning rate accordingly? Any advice will be much appreciated.

hello, which version of cuda are you using?

8.0. Which version should I use?

I don`t know, I encountered the same problem as you。I just want to confirm whether our environment is consistent。

Feb 06 '20 05:02 laojiangwei

@verymadmatt There might be some bugs in the pushed code and @hsfzxjy will check the possible reasons and update the progress soon. Please be patient.

Feb 06 '20 09:02 PkuRainBow

@verymadmatt We recommend you to try our "HRNet + OCR" on the other two datasets including PASCAL-Context and LIP.

Feb 08 '20 05:02 PkuRainBow

@verymadmatt We recommend you to try our "HRNet + OCR" on the other two datasets including PASCAL-Context and LIP.

Thanks for your response. So there're no bugs in the current release?

Feb 08 '20 09:02 verymadmatt

@verymadmatt Yes, the performance on Cityscapes is not very stable and we recommend you to run multiple times currently. The performance on the other datasets is expected to be more stable.

Feb 09 '20 05:02 PkuRainBow

@verymadmatt We perform class-balance for all the experiments on Cityscapes. We only use the Cityscapes train set for training. First, could you provide more details about your environmental information? We expect you to conduct experiments with Pytorch1.1. Second, please ensure that you have reproduced the performance of the HRNet baseline.

I read the code. But it seems that the parameter of class-balance is useless. In the 200th row of train.py, criterion = CrossEntropy(ignore_label=config.TRAIN.IGNORE_LABEL, weight=train_dataset.class_weights), so in the cityscapes dataset, class_weights is always use. I'm confused.

Dec 01 '20 07:12 purse1996

@purse1996 I could not quite understand what do you mean by "useless". The code snippet you post just shows that we are using class balance.

Dec 01 '20 08:12 hsfzxjy

Sorry, I did not express clearly what I mean. Of course, class weight is useful. While no matter that LOSS.CLASS_BALANCE is True or False, class-balance weights are always in use in criterion = CrossEntropy(ignore_label=config.TRAIN.IGNORE_LABEL, weight=train_dataset.class_weights). So I'm confused what is the use to turn the parameter of class_balance on? In this issue, https://github.com/HRNet/HRNet-Semantic-Segmentation/issues/91#issuecomment-579128770, he says "The mIoU increased to 0.8064 when I turned class_balance on". I'm confused about it. Thank you.

Dec 01 '20 08:12 purse1996

The config multi-scale and flip set as NO both in training and testing or training is YES, testing is NO?

Mar 17 '21 07:03 Margrate

HRNet-Semantic-Segmentation HRNet-Semantic-Segmentation copied to clipboard

performance gap of HRNetV2+OCR on cityscape val set using default config

HRNet-Semantic-Segmentation
HRNet-Semantic-Segmentation copied to clipboard