Bug in accuerate mode

[latot]

Hi hi again :3

Now I test the fast mode and accurate mode, while the first one works great, the accurate mode close the app when we try to run it, so to get the log, I run it from backend/main.py, here the log:

Running Verify Fluid Program ... 
Your Paddle Fluid works well on SINGLE GPU or CPU.
W1220 14:03:37.962996 14833 fuse_all_reduce_op_pass.cc:76] Find all_reduce operators: 2. To make the speed faster, some all_reduce ops are fused during training, after fusion, the number of all_reduce ops is 1.
Your Paddle Fluid works well on MUTIPLE GPU or CPU.
Your Paddle Fluid is installed successfully! Let's start deep Learning with Paddle Fluid now
识别字幕语言：ch
识别模式：accurate
请输入视频完整路径：~/video.mp4                                    
请输入字幕区域(ymin ymax xmin xmax)：0 720 0 1280
usage: main.py [-h] [--use_gpu USE_GPU] [--ir_optim IR_OPTIM] [--use_tensorrt USE_TENSORRT] [--use_fp16 USE_FP16] [--gpu_mem GPU_MEM]
               [--image_dir IMAGE_DIR] [--det_algorithm DET_ALGORITHM] [--det_model_dir DET_MODEL_DIR] [--det_limit_side_len DET_LIMIT_SIDE_LEN]
               [--det_limit_type DET_LIMIT_TYPE] [--det_db_thresh DET_DB_THRESH] [--det_db_box_thresh DET_DB_BOX_THRESH]
               [--det_db_unclip_ratio DET_DB_UNCLIP_RATIO] [--max_batch_size MAX_BATCH_SIZE] [--use_dilation USE_DILATION]
               [--det_east_score_thresh DET_EAST_SCORE_THRESH] [--det_east_cover_thresh DET_EAST_COVER_THRESH]
               [--det_east_nms_thresh DET_EAST_NMS_THRESH] [--det_sast_score_thresh DET_SAST_SCORE_THRESH]
               [--det_sast_nms_thresh DET_SAST_NMS_THRESH] [--det_sast_polygon DET_SAST_POLYGON] [--rec_algorithm REC_ALGORITHM]
               [--rec_model_dir REC_MODEL_DIR] [--rec_image_shape REC_IMAGE_SHAPE] [--rec_char_type REC_CHAR_TYPE] [--rec_batch_num REC_BATCH_NUM]
               [--max_text_length MAX_TEXT_LENGTH] [--rec_char_dict_path REC_CHAR_DICT_PATH] [--use_space_char USE_SPACE_CHAR]
               [--vis_font_path VIS_FONT_PATH] [--drop_score DROP_SCORE] [--use_angle_cls USE_ANGLE_CLS] [--cls_model_dir CLS_MODEL_DIR]
               [--cls_image_shape CLS_IMAGE_SHAPE] [--label_list LABEL_LIST] [--cls_batch_num CLS_BATCH_NUM] [--cls_thresh CLS_THRESH]
               [--enable_mkldnn ENABLE_MKLDNN] [--use_pdserving USE_PDSERVING]

optional arguments:
  -h, --help            show this help message and exit
  --use_gpu USE_GPU
  --ir_optim IR_OPTIM
  --use_tensorrt USE_TENSORRT
  --use_fp16 USE_FP16
  --gpu_mem GPU_MEM
  --image_dir IMAGE_DIR
  --det_algorithm DET_ALGORITHM
  --det_model_dir DET_MODEL_DIR
  --det_limit_side_len DET_LIMIT_SIDE_LEN
  --det_limit_type DET_LIMIT_TYPE
  --det_db_thresh DET_DB_THRESH
  --det_db_box_thresh DET_DB_BOX_THRESH
  --det_db_unclip_ratio DET_DB_UNCLIP_RATIO
  --max_batch_size MAX_BATCH_SIZE
  --use_dilation USE_DILATION
  --det_east_score_thresh DET_EAST_SCORE_THRESH
  --det_east_cover_thresh DET_EAST_COVER_THRESH
  --det_east_nms_thresh DET_EAST_NMS_THRESH
  --det_sast_score_thresh DET_SAST_SCORE_THRESH
  --det_sast_nms_thresh DET_SAST_NMS_THRESH
  --det_sast_polygon DET_SAST_POLYGON
  --rec_algorithm REC_ALGORITHM
  --rec_model_dir REC_MODEL_DIR
  --rec_image_shape REC_IMAGE_SHAPE
  --rec_char_type REC_CHAR_TYPE
  --rec_batch_num REC_BATCH_NUM
  --max_text_length MAX_TEXT_LENGTH
  --rec_char_dict_path REC_CHAR_DICT_PATH
  --use_space_char USE_SPACE_CHAR
  --vis_font_path VIS_FONT_PATH
  --drop_score DROP_SCORE
  --use_angle_cls USE_ANGLE_CLS
  --cls_model_dir CLS_MODEL_DIR
  --cls_image_shape CLS_IMAGE_SHAPE
  --label_list LABEL_LIST
  --cls_batch_num CLS_BATCH_NUM
  --cls_thresh CLS_THRESH
  --enable_mkldnn ENABLE_MKLDNN
  --use_pdserving USE_PDSERVING

The error, is very similar with what I reach to see in the gui, in the console case, we can change ~/video.mp4 to /home/user/video.mp4 and works, seems the project need to transform the input to full path, and the gui is not doing that, a faster way, would be the main.py transform the input to full path, that should fix this and enable relative paths.

[YaoFANGUK]

i have no idea how to fix that

[latot]

I check the code fast, sadly I don't get where the bug is, but there is a workaround that can do this works for now:

video_path = input(f"{interface_config['Main']['InputVideo']}").strip()

to

import os
video_path = os.path.abspath(input(f"{interface_config['Main']['InputVideo']}").strip())

This does not fix anything, so, this should be not be closed.

[YaoFANGUK]

i will test this tomorrow. Its time for sleep in China now.

[latot]

Okis!, good night :3