Skip to content

FCOS (Fully Convolutional One-Stage Object Detection) implementation in Keras and Tensorflow

Notifications You must be signed in to change notification settings

obiii/keras-fcos

 
 

Repository files navigation

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

image1 image2 image3

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.

About

FCOS (Fully Convolutional One-Stage Object Detection) implementation in Keras and Tensorflow

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%