This project is mainly based on faster_rcnn_pytorch.

For details about FPN please refer to the paper. Feature Pyramid Networks for Object Detection

• faster_rcnn_pytorch
• FPN
• Mask_RCNN

This project use ResNet-101 model converted from Caffe, and you can get it following RuotianLuo-pytorch-ResNet.

Since the program loading the data in pytorch_FPN/data by default, you can set the data path as following.

cd pytorch_FPN
mkdir data
cd data
ln -s $VOCdevkit VOCdevkit2007

Then you can set some hyper-parameters in train.py and training parameters in the .yml file.

python train.py

Set the path of the trained model in test.py.

cd pytorch_FPN
python demo.py

As FPN pointed out, anchor sizes for COCO may not suitable for VOC data, you can change anchor size to get better performance.

License: MIT license (MIT)