| Training Techniques | SGD with Momentum, Weight Decay |
|---|---|
| Architecture | 1x1 Convolution, Efficient Channel Attention, Bottleneck Residual Block, Batch Normalization, Convolution, Global Average Pooling, Residual Block, Residual Connection, ReLU, Max Pooling, Softmax, Squeeze-and-Excitation Block |
| ID | ecaresnet101d |
| SHOW MORE |
| Training Techniques | SGD with Momentum, Weight Decay |
|---|---|
| Architecture | 1x1 Convolution, Efficient Channel Attention, Bottleneck Residual Block, Batch Normalization, Convolution, Global Average Pooling, Residual Block, Residual Connection, ReLU, Max Pooling, Softmax, Squeeze-and-Excitation Block |
| ID | ecaresnet101d_pruned |
| SHOW MORE |
| Training Techniques | SGD with Momentum, Weight Decay |
|---|---|
| Architecture | 1x1 Convolution, Efficient Channel Attention, Bottleneck Residual Block, Batch Normalization, Convolution, Global Average Pooling, Residual Block, Residual Connection, ReLU, Max Pooling, Softmax, Squeeze-and-Excitation Block |
| ID | ecaresnet50d |
| SHOW MORE |
| Training Techniques | SGD with Momentum, Weight Decay |
|---|---|
| Architecture | 1x1 Convolution, Efficient Channel Attention, Bottleneck Residual Block, Batch Normalization, Convolution, Global Average Pooling, Residual Block, Residual Connection, ReLU, Max Pooling, Softmax, Squeeze-and-Excitation Block |
| ID | ecaresnet50d_pruned |
| SHOW MORE |
| Training Techniques | SGD with Momentum, Weight Decay |
|---|---|
| Architecture | 1x1 Convolution, Efficient Channel Attention, Bottleneck Residual Block, Batch Normalization, Convolution, Global Average Pooling, Residual Block, Residual Connection, ReLU, Max Pooling, Softmax, Squeeze-and-Excitation Block |
| ID | ecaresnetlight |
| SHOW MORE |
An ECA ResNet is a variant on a ResNet that utilises an Efficient Channel Attention module. Efficient Channel Attention is an architectural unit based on squeeze-and-excitation blocks that reduces model complexity without dimensionality reduction.
To load a pretrained model:
import timm
m = timm.create_model('ecaresnet50d', pretrained=True)
m.eval()
Replace the model name with the variant you want to use, e.g. ecaresnet50d. You can find the IDs in the model summaries at the top of this page.
You can follow the timm recipe scripts for training a new model afresh.
@misc{wang2020ecanet,
title={ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks},
author={Qilong Wang and Banggu Wu and Pengfei Zhu and Peihua Li and Wangmeng Zuo and Qinghua Hu},
year={2020},
eprint={1910.03151},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
Image Classification on ImageNet
| MODEL | TOP 1 ACCURACY | TOP 5 ACCURACY |
|---|---|---|
| ecaresnet101d | 82.18% | 96.06% |
| ecaresnet101d_pruned | 80.82% | 95.64% |
| ecaresnet50d | 80.61% | 95.31% |
| ecaresnetlight | 80.46% | 95.25% |
| ecaresnet50d_pruned | 79.71% | 94.88% |
