search

argusswift / YOLOv4-pytorch

This is a pytorch repository of YOLOv4, attentive YOLOv4 and mobilenet YOLOv4 with PASCAL VOC and COCO
1.68k stars 329 forks source link

why that my eval mAP is rather low? #204

Open sddai opened 3 years ago

sddai commented 3 years ago

i used the shared weight file and tested on voc, why the mAP is so low? `[2021-06-15 14:53:26,014]-[eval_voc.py line:91]:***Start Evaluation**** 100%|███████████████████████████████████████████████████████████████████████████████████████████████████| 4952/4952 [17:13<00:00, 4.79it/s] [2021-06-15 15:11:31,327]-[eval_voc.py line:99]:aeroplane --> mAP : 0.27931995093511486 [2021-06-15 15:11:31,327]-[eval_voc.py line:99]:bicycle --> mAP : 0.4199235817323248 [2021-06-15 15:11:31,328]-[eval_voc.py line:99]:bird --> mAP : 0.11354368432980091 [2021-06-15 15:11:31,328]-[eval_voc.py line:99]:boat --> mAP : 0.12539385468348616 [2021-06-15 15:11:31,328]-[eval_voc.py line:99]:bottle --> mAP : 0.0940665161769165 [2021-06-15 15:11:31,328]-[eval_voc.py line:99]:bus --> mAP : 0.5426845613815663 [2021-06-15 15:11:31,329]-[eval_voc.py line:99]:car --> mAP : 0.6271108491580774 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:cat --> mAP : 0.34451324857029475 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:chair --> mAP : 0.19507877793824396 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:cow --> mAP : 0.3548116302163225 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:diningtable --> mAP : 0.31734622887406977 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:dog --> mAP : 0.2806604323700429 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:horse --> mAP : 0.4753992133423637 [2021-06-15 15:11:31,331]-[eval_voc.py line:99]:motorbike --> mAP : 0.5036663806077839 [2021-06-15 15:11:31,332]-[eval_voc.py line:99]:person --> mAP : 0.6065566140452703 [2021-06-15 15:11:31,332]-[eval_voc.py line:99]:pottedplant --> mAP : 0.1607953530651103 [2021-06-15 15:11:31,333]-[eval_voc.py line:99]:sheep --> mAP : 0.04314534439258962 [2021-06-15 15:11:31,335]-[eval_voc.py line:99]:sofa --> mAP : 0.21280818412742938 [2021-06-15 15:11:31,335]-[eval_voc.py line:99]:train --> mAP : 0.38811412209232843 [2021-06-15 15:11:31,335]-[eval_voc.py line:99]:tvmonitor --> mAP : 0.4191048407322027

[2021-06-15 15:11:31,335]-[eval_voc.py line:103]:inference time: 2231.79 ms [2021-06-15 15:11:31,336]-[eval_voc.py line:105]: ===val cost time:1085.3216s `

argusswift commented 3 years ago

i used the shared weight file and tested on voc, why the mAP is so low? [2021-06-15 14:53:26,014]-[eval_voc.py line:91]:***********Start Evaluation**************** 100%|███████████████████████████████████████████████████████████████████████████████████████████████████| 4952/4952 [17:13<00:00, 4.79it/s] [2021-06-15 15:11:31,327]-[eval_voc.py line:99]:aeroplane --> mAP : 0.27931995093511486 [2021-06-15 15:11:31,327]-[eval_voc.py line:99]:bicycle --> mAP : 0.4199235817323248 [2021-06-15 15:11:31,328]-[eval_voc.py line:99]:bird --> mAP : 0.11354368432980091 [2021-06-15 15:11:31,328]-[eval_voc.py line:99]:boat --> mAP : 0.12539385468348616 [2021-06-15 15:11:31,328]-[eval_voc.py line:99]:bottle --> mAP : 0.0940665161769165 [2021-06-15 15:11:31,328]-[eval_voc.py line:99]:bus --> mAP : 0.5426845613815663 [2021-06-15 15:11:31,329]-[eval_voc.py line:99]:car --> mAP : 0.6271108491580774 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:cat --> mAP : 0.34451324857029475 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:chair --> mAP : 0.19507877793824396 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:cow --> mAP : 0.3548116302163225 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:diningtable --> mAP : 0.31734622887406977 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:dog --> mAP : 0.2806604323700429 [2021-06-15 15:11:31,330]-[eval_voc.py line:99]:horse --> mAP : 0.4753992133423637 [2021-06-15 15:11:31,331]-[eval_voc.py line:99]:motorbike --> mAP : 0.5036663806077839 [2021-06-15 15:11:31,332]-[eval_voc.py line:99]:person --> mAP : 0.6065566140452703 [2021-06-15 15:11:31,332]-[eval_voc.py line:99]:pottedplant --> mAP : 0.1607953530651103 [2021-06-15 15:11:31,333]-[eval_voc.py line:99]:sheep --> mAP : 0.04314534439258962 [2021-06-15 15:11:31,335]-[eval_voc.py line:99]:sofa --> mAP : 0.21280818412742938 [2021-06-15 15:11:31,335]-[eval_voc.py line:99]:train --> mAP : 0.38811412209232843 [2021-06-15 15:11:31,335]-[eval_voc.py line:99]:tvmonitor --> mAP : 0.4191048407322027 [2021-06-15 15:11:31,335]-[eval_voc.py line:102]:mAP:0.325202168438567 [2021-06-15 15:11:31,335]-[eval_voc.py line:103]:inference time: 2231.79 ms [2021-06-15 15:11:31,336]-[eval_voc.py line:105]: ===val cost time:1085.3216s

已修正