PyTorch implementation of Deformable Convolution

!!!Warning: There is some issues in this implementation and this repo is not maintained any more, please consider using for example: TORCHVISION.OPS.DEFORM_CONV

• By Wei OUYANG @ Institut Pasteur
• Thanks to Felix Lau's Keras/TensorFlow implementation: ~~https://github.com/felixlaumon/deform-conv~~ (https://github.com/kastnerkyle/deform-conv)

TODO List

• [x] implement offsets mapping in pytorch
• [x] all tests passed
• [x] deformable convolution module
• [x] Fine-tuning the deformable convolution modules
• [x] scaled mnist demo
• [x] improve speed with cached grid array
• [ ] use MNIST dataset from pytorch (instead of Keras)
• [ ] support input image with different width and height
• [ ] benchmark with tensorflow implementation

Deformable Convolutional Networks

Dai, Jifeng, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, and Yichen Wei. 2017. “Deformable Convolutional Networks.” arXiv [cs.CV]. arXiv. http://arxiv.org/abs/1703.06211

The following animation is generated by Felix Lau (with his tensorflow implementation):

Also Check out Felix Lau's summary of the paper: https://medium.com/@phelixlau/notes-on-deformable-convolutional-networks-baaabbc11cf3