This is an implementation of RetinaNet in Pytorch, using ResNet as backbone and FPN. Its base on the code of Detectron and pytorch-retinanet.

1.Download PASCAL VOC 2012 trainval datasets and unzip it. Its path should be '{root_dir}/VOCdevkit/..'

2.Download this repo

git clone [email protected]:qqadssp/RetinaNet.git  
cd RetinaNet  

3.Download pretrained weights from https://download.pytorch.org/models/resnet50-19c8e357.pth

cd checkpoint  
wget https://download.pythorch.org/models/resnet50-19c8e357.pth  
cd ..  

4.Initialize the model

python init.py  

5.Modify configs file in 'config'. For VOC datatsets, modify 'TRAIN: DATASETS_DIR' with your {root_dir}

6.Trian the model

python train.py --cfg ./configs/RetinaNet_ResNet50_FON_VOC.yaml  

1.Download PASCAL VOC 2012 test datasets and unzip it. Its path should be '{root_dir}/VOCdevkit_test/..'

2.Modify ocnfig file in 'configs'. For VOC datasets, modify 'TEST: DATASETS_DIR' with your {root_dir}, and 'TEST: WEIGHTS' with the trained weights in 'checkpoint'

3.Test the model. The result files will be in 'result'.

python test.py --cfg ./configs/RetinaNet_ResNet50_FPN_VOC.yaml