PyTorch_YOLOv4
PyTorch_YOLOv4 copied to clipboard
WongKinYiu
PyTorch implementation of YOLOv4
 Scanning labels D:\Anaconda3\envs\pytorch\ayolo\pytorch_YOLOv4\VOCdevkit\VOC2019\JPEGImages.cache (864 found, 0 missing, 0 empty, 0 duplicate, for 864 images): 100%|█| 864/864 Class Images Targets P R [email protected] [email protected]:.95: 100%|███████████████████████████████████████████| 216/216 [02:43
Show as the ScaledYOLOv4, the coco val is ``` Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.48656 Average Precision (AP) @[ IoU=0.50 | area= all...
I saw you development log "2021-07-22 - support 1) decoupled head, 2) anchor-free, and 3) multi positives" I have a question: can the decoupled head be used for the anchor-based...
把训练好的模型.pt转换成.weight，和.cfg .names用到darknet里 测出来的结果和用.pt测的结果相差很大 请问我需要做什么修改吗？ when I use the .weight file changed from trained .pt file in darknet there is difference of the results. do I need to change something? thank...
``` for custom dataset, usually i will split the dataset into development set (training set + validation set) and testing set. training set is used to training and validation set...
@WongKinYiu 您好: 想請教有提供backbone pretrained model嗎? 例如我想訓練自己的detector(yolo layer), 但是backbone(cspdarknet53)的部分不想重新訓練. 我對object detection models的理解是會先用backbone在ImageNet 上training完成, 再與detector 整合end to end 在MSCOCO上做training 謝謝
Hello, I use DDP mode to find that the training time of two GPUs is 26min and that of one GPU is 16min in an epoch. Do you know why?...
darknet版本的yolov4在输入为608*608时，map是43.5%，这个版本在输入为640*640时map是50%。在输入尺寸相近情况下，请问为什么差距这么大？
How can I use pre trained weights on COCO dataset and finetune them on my dataset for yolov4 and yolov4 tiny? While loading the model only Pytorch weights are used....
the parameters of banch and minibanch are set in the yolov4.cfg file, why also use --banch-size ' ' in command line when running program? 
