Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/hujie-frank/SENet

Squeeze-and-Excitation Networks
https://github.com/hujie-frank/SENet

caffe gpu senet

Last synced: about 2 months ago
JSON representation

Squeeze-and-Excitation Networks

Awesome Lists containing this project

README 

        
# Squeeze-and-Excitation Networks ([paper](http://openaccess.thecvf.com/content_cvpr_2018/papers/Hu_Squeeze-and-Excitation_Networks_CVPR_2018_paper.pdf))

By Jie Hu[1], Li Shen[2], Gang Sun[1].



[Momenta](https://momenta.ai/)[1] and [University of Oxford](http://www.robots.ox.ac.uk/~vgg/)[2].



## Approach



  





  Figure 1: Diagram of a Squeeze-and-Excitation building block.






   

  





  Figure 2: Schema of SE-Inception and SE-ResNet modules. We set r=16 in all our models.




## Implementation

In this repository, Squeeze-and-Excitation Networks are implemented by [Caffe](https://github.com/BVLC/caffe).



### Augmentation

| Method | Settings |

|:-:|:-:|

|Random Mirror| True |

|Random Crop| 8% ~ 100% |

|Aspect Ratio | 3/4 ~ 4/3 |

|Random Rotation| -10° ~ 10°|

|Pixel Jitter| -20 ~ 20 |



### Note:

* To achieve efficient training and testing, we combine the consecutive operations ***channel-wise scale*** and ***element-wise summation*** into a single layer **"Axpy"** in the architectures with skip-connections, resulting in a considerable reduction in memory cost and computational burden.



* In addition, we found that the implementation for ***global average pooling*** on GPU supported by cuDNN and BVLC/caffe is less efficient. In this regard, we re-implement the operation which achieves significant acceleration.



## Trained Models



Table 1. Single crop validation error on ImageNet-1k (center 224x224 crop from resized image with shorter side = 256). The SENet-154 is one of our superior models used in [ILSVRC 2017 Image Classification Challenge](http://image-net.org/challenges/LSVRC/2017/index) where we won the 1st place (Team name: [WMW](http://image-net.org/challenges/LSVRC/2017/results)).



| Model | Top-1 | Top-5 | Size | Caffe Model | Caffe Model

|:-:|:-:|:-:|:-:|:-:|:-:|

|SE-BN-Inception| 23.62 | 7.04 | 46 M| [GoogleDrive](https://drive.google.com/file/d/0BwHV3BlNKkWlTWRRbDZYbVB2WWc/view?usp=sharing) | [BaiduYun](https://pan.baidu.com/s/1qYoPdak)

|SE-ResNet-50   | 22.37 | 6.36 | 107 M | [GoogleDrive](https://drive.google.com/file/d/0BwHV3BlNKkWlS2QwZHFzM3RjNzg/view?usp=sharing) | [BaiduYun](https://pan.baidu.com/s/1gf5wsLl)

|SE-ResNet-101  | 21.75  | 5.72 | 189 M | [GoogleDrive](https://drive.google.com/file/d/0BwHV3BlNKkWlTEg4YmcwQ0FoZFU/view?usp=sharing) | [BaiduYun](https://pan.baidu.com/s/1c1FvCWg)

|SE-ResNet-152  | 21.34  | 5.54 | 256 M | [GoogleDrive](https://drive.google.com/file/d/0BwHV3BlNKkWlcFE0Q2NTcWl3WUE/view?usp=sharing) | [BaiduYun](https://pan.baidu.com/s/1dFEnSzR)

|SE-ResNeXt-50 (32 x 4d) | 20.97 | 5.54 | 105 M | [GoogleDrive](https://drive.google.com/file/d/0BwHV3BlNKkWlQ2Z0Q204V1RITjA/view?usp=sharing) | [BaiduYun](https://pan.baidu.com/s/1dFbEmbv)

|SE-ResNeXt-101 (32 x 4d) | 19.81 | 4.96 | 187 M | [GoogleDrive](https://drive.google.com/file/d/0BwHV3BlNKkWleklsNzBiZlprblk/view?usp=sharing) | [BaiduYun](https://pan.baidu.com/s/1qY2wjt6)

|SENet-154 | 18.68 | 4.47 | 440 M | [GoogleDrive](https://drive.google.com/file/d/0BwHV3BlNKkWlbTFZbzFTSXBUTUE/view?usp=sharing) | [BaiduYun](https://pan.baidu.com/s/1o7HdfAE)



Here we obtain better performance than those reported in the paper.

We re-train the SENets described in the paper on a single GPU server with 8 NVIDIA Titan X cards, using a mini-batch of 256 and a initial learning rate of 0.1 with more epoches. 

In contrast, the results reported in the paper were obtained by training the networks with a larger batch size (1024) and learning rate (0.6) across 4 servers. 



## Third-party re-implementations

0. Caffe. SE-mudolues are integrated with a modificated ResNet-50 using a stride 2 in the 3x3 convolution instead of the first 1x1 convolution which obtains better performance: [Repository](https://github.com/shicai/SENet-Caffe).

0. TensorFlow. SE-modules are integrated with a pre-activation ResNet-50 which follows the setup in [fb.resnet.torch](https://github.com/facebook/fb.resnet.torch): [Repository](https://github.com/ppwwyyxx/tensorpack/tree/master/examples/ResNet).

0. TensorFlow. Simple Tensorflow implementation of SENets using Cifar10: [Repository](https://github.com/taki0112/SENet-Tensorflow).

0. MatConvNet. All the released SENets are imported into [MatConvNet](https://github.com/vlfeat/matconvnet): [Repository](https://github.com/albanie/mcnSENets).

0. MXNet. SE-modules are integrated with the ResNeXt and more architectures are coming soon: [Repository](https://github.com/bruinxiong/SENet.mxnet).

0. PyTorch. Implementation of SENets by PyTorch: [Repository](https://github.com/moskomule/senet.pytorch).

0. Chainer. Implementation of SENets by Chainer: [Repository](https://github.com/nutszebra/SENets).

## Citation



If you use Squeeze-and-Excitation Networks in your research, please cite the paper:

    

    @inproceedings{hu2018senet,

      title={Squeeze-and-Excitation Networks},

      author={Jie Hu and Li Shen and Gang Sun},

      journal={IEEE Conference on Computer Vision and Pattern Recognition},

      year={2018}

    }