RetinaNet: Focal Loss for Dense Object Detection

this repo's code is similar with my another repo : https://github.com/ChingHo97/FCOS-PyTorch-37.2AP you can see difference between anchor-base and anchor-free

AP Result

PASCAL VOC (800px)	COCO(800px)
81.6 (IoU.5)	36.4

Requirements

• opencv-python
• pytorch >= 1.0
• torchvision >= 0.4.
• matplotlib
• cython
• numpy == 1.17
• Pillow
• tqdm
• pycocotools

Results in coco

Train coco2017 on 4 Tesla-V100, 4 imgs for each gpu, init lr=1e-2 using GN GIou.

You can download the 36.4 ap result in Baidu driver link, password: 421x,then put it in checkpoint folder, then run the coco_eval_retina.py

Results in Pascal Voc

Train Voc07+12 on 4 Tesla-V100 , 4 imgs for each gpu, init lr=1e-2 using GN,GIou.

You can download the 81.6 ap result in Baidu driver link, password:emkw, then put it in checkpoint folder, then run the eval_voc_retina.py and

train for coco

You can run the train_coco_retina.py, train 24 epoch and you can get the result. You need to change the coco2017 path.

train for PASCAL VOC

You can run the train_voc_retina.py, train 30 epoch and you can get the result. You need to change the PASCAL07+12 path, you can reference to this repo:https://github.com/YuwenXiong/py-R-FCN

Detect Image

You can run the detect.py to detect images , this repo provides PASCAL VOC Images detection demo.