/SENet

Squeeze-and-Excitation Networks

Primary LanguageCudaApache License 2.0Apache-2.0

Squeeze-and-Excitation Networks (paper)

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

Momenta[1] and University of Oxford[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.

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 where we won the 1st place (Team name: WMW).

Model Top-1 Top-5 Size Caffe Model Caffe Model
SE-BN-Inception 23.62 7.04 46 M GoogleDrive BaiduYun
SE-ResNet-50 22.37 6.36 107 M GoogleDrive BaiduYun
SE-ResNet-101 21.75 5.72 189 M GoogleDrive BaiduYun
SE-ResNet-152 21.34 5.54 256 M GoogleDrive BaiduYun
SE-ResNeXt-50 (32 x 4d) 20.97 5.54 105 M GoogleDrive BaiduYun
SE-ResNeXt-101 (32 x 4d) 19.81 4.96 187 M GoogleDrive BaiduYun
SENet-154 18.68 4.47 440 M GoogleDrive BaiduYun

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

  1. 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.
  2. TensorFlow. SE-modules are integrated with a pre-activation ResNet-50 which follows the setup in fb.resnet.torch: Repository.
  3. TensorFlow. Simple Tensorflow implementation of SENets using Cifar10: Repository.
  4. MatConvNet. All the released SENets are imported into MatConvNet: Repository.
  5. MXNet. SE-modules are integrated with the ResNeXt and more architectures are coming soon: Repository.
  6. PyTorch. Implementation of SENets by PyTorch: Repository.
  7. Chainer. Implementation of SENets by Chainer: Repository.

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}
}