https://github.com/oeway/pytorch-deform-conv

PyTorch implementation of Deformable Convolution
https://github.com/oeway/pytorch-deform-conv

deep-learning deep-learning-algorithms

Last synced: about 2 months ago
JSON representation

PyTorch implementation of Deformable Convolution

• Host: GitHub
• URL: https://github.com/oeway/pytorch-deform-conv
• Owner: oeway
• License: mit
• Archived: true
• Created: 2017-04-04T18:32:54.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2021-07-21T12:48:04.000Z (almost 3 years ago)
• Last Synced: 2024-01-25T20:08:55.703Z (4 months ago)
• Topics: deep-learning, deep-learning-algorithms
• Language: Python
• Size: 1.49 MB
• Stars: 903
• Watchers: 21
• Forks: 149
• Open Issues: 15
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Lists

• Awesome-pytorch-list - pytorch-deform-conv
• my-awesome-stars - oeway/pytorch-deform-conv - PyTorch implementation of Deformable Convolution (Python)
• awesome-pytorch - Deformable Convolution
• Awesome-pytorch-list-CNVersion - pytorch-deform-conv
• awesome-pytorch - Deformable Convolution

README

        
# 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](https://pytorch.org/vision/stable/_modules/torchvision/ops/deform_conv.html)**

 

 * 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):

![](deformable-learned-offset-filtered.gif)

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