Ecosyste.ms: Awesome

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

Awesome Lists | Featured Topics | Projects

https://github.com/implus/SKNet

Code for our CVPR 2019 paper: Selective Kernel Networks; See zhihu：https://zhuanlan.zhihu.com/p/59690223
https://github.com/implus/SKNet

cnn selective-kernel-networks sknet

Last synced: about 1 month ago
JSON representation

Code for our CVPR 2019 paper: Selective Kernel Networks; See zhihu：https://zhuanlan.zhihu.com/p/59690223

Awesome Lists containing this project

README 

        
# SKNet: Selective Kernel Networks ([paper](https://arxiv.org/pdf/1903.06586.pdf))

By Xiang Li[1,2], Wenhai Wang[3,2], Xiaolin Hu[4] and Jian Yang[1]



[PCALab, Nanjing University of Science and Technology][1] [Momenta](https://momenta.ai/)[2] [Nanjing University][3] [Tsinghua University][4].



## Approach



  





  Figure 1: The Diagram of a Selective Kernel Convolution module.




## Implementation

In this repository, all the models are implemented by [Caffe](https://github.com/BVLC/caffe).

 

We use the data augmentation strategies with [SENet](https://github.com/hujie-frank/SENet). 



There are two new layers introduced for efficient training and inference, these are *Axpy* and *CuDNNBatchNorm* layers.  

+ The [*Axpy*](https://github.com/hujie-frank/SENet/blob/master/src/caffe/layers/) layer is already implemented in [SENet](https://github.com/hujie-frank/SENet).

+ The [*CuDNNBatchNorm*] is mainly borrowed from [GENet](https://github.com/hujie-frank/GENet).



## Trained Models

Table 2. Single crop validation error on ImageNet-1k (center 224x224/320x320 crop from resized image with shorter side = 256). 



| Model | Top-1 224x | Top-1 320x | #P | GFLOPs | 

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

|ResNeXt-50        |22.23|21.05|25.0M|4.24|

|AttentionNeXt-56  |21.76|–    |31.9M|6.32|

|InceptionV3       |–    |21.20|27.1M|5.73|

|ResNeXt-50 + BAM  |21.70|20.15|25.4M|4.31|

|ResNeXt-50 + CBAM |21.40|20.38|27.7M|4.25|

|SENet-50          |21.12|19.71|27.7M|4.25|

|SKNet-50          |20.79|19.32|27.5M|4.47|

|ResNeXt-101       |21.11|19.86|44.3M|7.99|

|Attention-92      | –   |19.50|51.3M|10.43|

|DPN-92            |20.70|19.30|37.7M|6.50|

|DPN-98            |20.20|18.90|61.6M|11.70|

|InceptionV4       | –   |20.00|42.0M|12.31|

|Inception-ResNetV2| –   |19.90|55.0M|13.22|

|ResNeXt-101 + BAM |20.67|19.15|44.6M|8.05|

|ResNeXt-101 + CBAM|20.60|19.42|49.2M|8.00|

|SENet-101         |20.58|18.61|49.2M|8.00|

|SKNet-101         |20.19|18.40|48.9M|8.46|



Download:



|Model|caffe model|

|:-:|:-:|

|SKNet-50|[GoogleDrive](https://drive.google.com/file/d/1EKanqFkqoU3L6vgSLW3GjPciesZ2rrUH/view?usp=sharing)|

|SKNet-101|[GoogleDrive](https://drive.google.com/file/d/1NEYIYeSXQeinLGU5GH2hf8T6MJytzt0U/view?usp=sharing)|



20190323_Update: SKNet-101 model is deleted by mistake. We are retraining a model and it will come soon in 2-3 days.

20190326_Update: SKNet-101 model is ready.



## Attention weights correspond to object scales in low/middle layers

We look deep into the selection distributions from the perspective of classes on SK_2_3 (low), SK_3_4 (middle), SK_5_3 (high) layers:



  





  Figure 2: Average mean attention difference (mean attention value of kernel 5x5 minus that of kernel 3x3) on SK units of SKNet-50, for each of 1,000 categories using all validation samples on ImageNet. On low or middle level SK units (e.g., SK\_2\_3, SK\_3\_4), 5x5 kernels are clearly imposed with more emphasis if the target object becomes larger (1.0x -> 1.5x).




More details of attention distributions on specific images are as follows:



  




## Citation



If you use Selective Kernel Convolution in your research, please cite the paper:

    

    @inproceedings{li2019selective,

      title={Selective Kernel Networks},

      author={Li, Xiang and Wang, Wenhai and Hu, Xiaolin and Yang, Jian},

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

      year={2019}

    }