search

tanluren / yolov3-channel-and-layer-pruning

yolov3 yolov4 channel and layer pruning, Knowledge Distillation 层剪枝，通道剪枝，知识蒸馏
Apache License 2.0
1.5k stars 446 forks source link

No results in evaluation P R and mAP are 0 #84

Open qaazii opened 4 years ago

qaazii commented 4 years ago

Hello @zbyuan Please check my am i getting the P R and mAP zero. Below is the values after each epoch 

 Epoch   gpu_mem      GIoU       obj       cls     total      soft    rratio   targets  img_size

0%| | 0/1355 [00:00<?, ?it/s]learning rate: 1e-06 0/2 7.65G 2.44 1.74 41.9 46.1 0 0 3 416: 100%|█| 1355/1355 Reading image shapes: 100%|███████████████████████████████████████████████████████| 5417/5417 [00:17<00:00, 303.63it/s] Class Images Targets P R mAP F1: 100%|█| 339/339 [01:13<00:00, 4.63it all 5.42e+03 6.14e+03 0 0 0 0

zbyuan commented 4 years ago

Load pre-training weights？Date problem？QQ group : 734912150

qaazii commented 4 years ago

Load pre-training weights？Date problem？QQ group : 734912150

I have this error when The code is evaluating after every epoch

zbyuan commented 4 years ago

Um, you can train yourself with dozens of data . Encountered the same problem？May be a data problem

tanluren commented 4 years ago

check the train_batch0.jpg and test_batch0.jpg to see if you prepair the data right

qaazii commented 4 years ago

check the train_batch0.jpg and test_batch0.jpg to see if you prepair the data right Hi @zbyuan @tanluren Is it possible that I talk to you on wechat Please. I can explain my problem.

Actually I want to train the teacher model on different classes and student model on different classes and in testing I wan to predict both.

zhuhuangru commented 4 years ago

Please check your dataset. These are some difference between officail YOLO dataset and ultralytics version Dataset.

zhangnan-hust commented 3 years ago

Would you please tell me how different between the official yolo dataset and ultralytics version dataset, and how to generate the dataset and labels? I met the same problem, and I used the voc_label.py to create the trainval.txt, test.txt which including the path of images and labels which including the classes and the information of bbox.

wenhaotang commented 3 years ago

Load pre-training weights？Date problem？QQ group : 734912150

Hi @zbyuan , the QQ group : 734912150 is not found now. How can I find the group? : )

zbyuan commented 3 years ago

@wenhaotang Your qq number

wenhaotang commented 3 years ago

@wenhaotang Your qq number

@zbyuan My qq number: 423184965 Thanks!