Mask RCNN

Mask RCNN in TensorFlow

This repo attempts to reproduce this amazing work by Kaiming He et al. : Mask R-CNN

Requirements

• Tensorflow (>= 1.0.0)
• Numpy
• COCO dataset
• Resnet50

How-to

1. Go to ./libs/datasets/pycocotools and run make
2. Download COCO dataset, place it into ./data, then run python download_and_convert_data.py to build tf-records. It takes a while.
3. Download pretrained resnet50 model, wget http://download.tensorflow.org/models/resnet_v1_50_2016_08_28.tar.gz, unzip it, place it into ./data/pretrained_models/
4. Go to ./libs and run make
5. run python train/train.py for training
6. There are certainly some bugs, please report them back, and let's solve them together.

TODO:

• [x] ROIAlign
• [x] COCO Data Provider
• [x] Resnet50
• [x] Feature Pyramid Network
• [x] Anchor and ROI layer
• [x] Mask layer
• [x] Speedup anchor layer with cython
• [x] Combining all modules together.
• [x] Testing and debugging (in progress)
• [ ] Training / evaluation on COCO
• [ ] Add image summary to show some results
• [ ] Converting ResneXt
• [ ] Training >2 images

Call for contributions

• Anything helps this repo, including discussion, testing, promotion and of course your awesome code.

Acknowledgment

This repo borrows tons of code from

• TFFRCNN
• py-faster-rcnn
• faster_rcnn
• tf-models

License

See LICENSE for details.