This is a Pytorch implementation of the paper "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks". I reference faster_rcnn_pytorch and faster rcnn pytorch tuturial pytorch pull request but this request is not completed and closed.

I was about to study faster rcnn code but codes are too difficult to me. I need some simple tutorial but there is no simple code. so I tried to write simple code than other repo.

this repo's code are

• pytoch
• tensorflow (Using tensorboad)
• matplotlib

I use floydhub to train model
Floydhub is simple deeplearining training tool

pip install -U floyd-cli

#./input
floyd data init voc
floyd data upload

#./FRCNN
floyd init frcnn
floyd data status
floyd run --env pytorch --gpu --data [your data id] "python3 main.py"

This porject structure is fitted with floydhub structure, so parent directory contain input, output, FRCNN directory

but you can traning any environment without floydhub

Follow this project (TFFRCNN) to download and prepare the training, validation, test data and the VGG16 model pre-trained on ImageNet.

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

1000 epochs

faster_rcnn_pytorch
pytorch pull request.