yolov3-pytroch:

Calculate the gradient value directly like Darknet.

	IOU	area	maxDets	yolov3 paper	this impl
Average Precision (AP)	IoU=0.50:0.95	all	100	0.33	0.343
Average Precision (AP)	IoU=0.50	all	100	0.579	0.572
Average Precision (AP)	IoU=0.75	all	100	0.344	0.365
Average Precision (AP)	IoU=0.50:0.95	small	100	0.183	0.181
Average Precision (AP)	IoU=0.50:0.95	medium	100	0.354	0.377
Average Precision (AP)	IoU=0.50:0.95	large	100	0.419	0.451

	IOU	area	maxDets	this impl
Average Precision (AP)	IoU=0.50:0.95	all	100	0.321
Average Precision (AP)	IoU=0.50	all	100	0.541
Average Precision (AP)	IoU=0.75	all	100	0.335
Average Precision (AP)	IoU=0.50:0.95	small	100	0.140
Average Precision (AP)	IoU=0.50:0.95	medium	100	0.349
Average Precision (AP)	IoU=0.50:0.95	large	100	0.478

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
cfg		cfg
dataset		dataset
lib		lib
log		log
mAP		mAP
model		model
eval_coco.py		eval_coco.py
eval_voc.py		eval_voc.py
readme.md		readme.md
test_coco.py		test_coco.py
train_coco.py		train_coco.py