keras-fcos

This is an implementation of FCOS on keras and Tensorflow. The project is based on fizyr/keras-retinanet and tianzhi0549/FCOS. Thanks for their hard work.

Test

1. I trained on Pascal VOC2012 trainval.txt + Pascal VOC2007 train.txt, and validated on Pascal VOC2007 val.txt. There are 14041 images for training and 2510 images for validation.
2. The best evaluation results on VOC2007 test are (score_threshold=0.05):

backbone	mAP50
resnet50	0.6892
resnet101	0.7352

3. Pretrained model is here. baidu netdisk extract code: yr8k
4. python3 inference.py to test your image by specifying image path and model path there.

Train

build dataset (Pascal VOC, other types please refer to fizyr/keras-retinanet)

• Download VOC2007 and VOC2012, copy all image files from VOC2007 to VOC2012.
• Append VOC2007 train.txt to VOC2012 trainval.txt.
• Overwrite VOC2012 val.txt by VOC2007 val.txt.

train

• python3 train.py --backbone resnet50 --gpu 0 pascal datasets/VOC2012 to start training.

Evaluate

• python3 utils/eval.py to evaluate by specifying model path there.