Deformable Convolutional Op Support

[rootkitchao]

System information

• TensorFlow version (you are using):2.0 alpha
• TensorFlow Addons version:0,21
• Is it in the tf.contrib (if so, where):no
• Are you willing to contribute it (yes/no):yes,if i know how to do that
• Are you willing to maintain it going forward? (yes/no):yes

Describe the feature and the current behavior/state.
Deformable Convolutional Network(https://arxiv.org/pdf/1703.06211.pdf,https://arxiv.org/pdf/1703.06211.pdf) has achieved good results in the du detection task.But the current version of Tensorflow does not seem to implement deformable convolution operations.The efficiency of some third-party implementations of deformable convolution operations is lower than that of the mxnet version(https://github.com/msracver/Deformable-ConvNets).I hope that Tensorflow can add native deformable convolution operations in future versions.
Thanks.
Will this change the current api? How?
Yes,Add a new op
Who will benefit with this feature?
Any one want to use tensorflow for object detection.
Any Other info.

[Squadrick]

Here's a python only implementation of v1 for Keras which can be ported to TF-Keras. This repository is the official implementation with CUDA code; it's written using MXNet but it can be ported over to TensorFlow.

Here are the papers:

1. Deformable Convolutional Networks
2. Deformable ConvNets v2

I can start working on this if needed. I'll get the python implementation for v2 up and running first, followed by the proper CUDA/C++ implementation.

[rootkitchao]

Here's a python only implementation of v1 for Keras which can be ported to TF-Keras. This repository is the official implementation with CUDA code; it's written using MXNet but it can be ported over to TensorFlow.
Here are the papers:

Deformable Convolutional Networks
Deformable ConvNets v2

I can start working on this if needed. I'll get the python implementation for v2 up and running first, followed by the proper CUDA/C++ implementation.

Thanks for reply.And here is a Mask RCNN implemented by Tensorflow that may help test.

[facaiy]

@rootkitchao Welcome, William. I think you can go first because you seems willing to contribute the codes, right?

Are you willing to contribute it (yes/no):yes,if i know how to do that

Sorry for the delay, and thank @Squadrick for the reference paper and implementation :-)

[Squadrick]

@rootkitchao Any update? I can work on this if you aren't.

[Squadrick]

Using native CUDA for deform conv is dependant on #118 .

[jlhervy]

I tried to look into it, but it is too hard for me to do as I don't have experience in CUDA.

I wonder how hard would it be to port this implementation (tf1.4) to tf2.0 :
https://github.com/Zardinality/TF-deformable-conv

This is the first version of the paper, but ideally it would be nice to have deformable-conv V2 :
https://arxiv.org/abs/1811.11168
For which there is an implementation for pytorch, as cited by @Squadrick

Thank you

[vlordier]

@Squadrick you mentioned you'd implement DCN v2 in TF 2 : is this something you're having progress with ?

[Squadrick]

@vlordier Nope, haven't found the time to get around to this. Feel free to open a PR if you decide to work on this.

[bhack]

/cc @smallsunsun1

[smallsunsun1]

/cc @smallsunsun1

I implemented a Deformable Conv2D V2 And DeformablePsROIAlign operation In both cpu and gpu according to mmdetection https://github.com/open-mmlab/mmdetection. My implenmentation is https://github.com/smallsunsun1/custom_ops/tree/master/deformable_conv2d_ops_new , which support tf2.0 and built by CMake.

[bhack]

@smallsunsun1 Do you want to officially PR your impl on TF addons?

[smallsunsun1]

@bhack Maybe I can have a try, but I haven't PR before, it may task some time==

[seanpmorgan]

@bhack Maybe I can have a try, but I haven't PR before, it may task some time==

Please let us know if there is anyway we can help. We have some documentation to get you started.

TFA pull requests: https://github.com/tensorflow/addons/blob/master/CONTRIBUTING.md#pull-requests
Getting started with contributions: https://github.com/gabrieldemarmiesse/getting_started_open_source

[smallsunsun1]

@seanpmorgan I clone the addons repo, tested my deformable_conv2d impl in local, building passed in tensorflow2.0 and cuda10.0, noticed the lastest branch is target in tensorflow2.1 and cuda10.1. But I can't have a cuda10.1 environment to test the code. Is there any problem if I push a request

[seanpmorgan]

@seanpmorgan I clone the addons repo, tested my deformable_conv2d impl in local, building passed in tensorflow2.0 and cuda10.0, noticed the lastest branch is target in tensorflow2.1 and cuda10.1. But I can't have a cuda10.1 environment to test the code. Is there any problem if I push a request

Not at all. We have CI testing that will run it on GPU for you (So even if you didn't have a GPU you could still test it.) We welcome WIP PRs so feel free to start the PR and we can go from there.

[smallsunsun1]

already open a PR #1129, support DeformableConv2d ，DeformablepsROIalign and EquiConv

[alexattia]

Do you have any news about the progress of this PR and deformable convolutions in TensorFlow?

[Licht-T]

I've created new PR since the previous PR is stale and contains the unknown license code. I'd like to make my PR reviewed before it gets stale ;-).
#2196

[seyed-mohammadreza-mousavi]

Hello. I hope you are good. There is no implementation of deformable convolution in tensorflow2. Could someone please help me?
There is a deform conv class where you can use it in tensorflow2 but you can not use it in tensorflow model?
How can I make this class to be usable in models?

[Cospel]

@aAmohammadrezaaA The state of deformable convolution in tensorflow 2 is tearful. There were several different merge requests with different implementations. I tried in the past some of them and all were quite good and working. However both tensorflow team and tf-addons team have little time to do good PR and integration it to the framework. The best advice right now is to try different framework if you want to have a good support of deformable convolutions.

[seanpmorgan]

TensorFlow Addons is transitioning to a minimal maintenance and release mode. New features will not be added to this repository. For more information, please see our public messaging on this decision:
TensorFlow Addons Wind Down

Please consider sending feature requests / contributions to other repositories in the TF community with a similar charters to TFA:
Keras
Keras-CV
Keras-NLP