Training Techniques | SGD with Momentum, Weight Decay |
---|---|
Architecture | Convolution, Residual Connection, Dense Connections, Global Average Pooling, Max Pooling, Selective Kernel, Softmax |
ID | skresnet18 |
SHOW MORE |
Training Techniques | SGD with Momentum, Weight Decay |
---|---|
Architecture | Convolution, Residual Connection, Dense Connections, Global Average Pooling, Max Pooling, Selective Kernel, Softmax |
ID | skresnet34 |
SHOW MORE |
SK ResNet is a variant of a ResNet that employs a Selective Kernel unit. In general, all the large kernel convolutions in the original bottleneck blocks in ResNet are replaced by the proposed SK convolutions, enabling the network to choose appropriate receptive field sizes in an adaptive manner.
To load a pretrained model:
import timm
m = timm.create_model('skresnet34', pretrained=True)
m.eval()
Replace the model name with the variant you want to use, e.g. skresnet34
. 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{li2019selective,
title={Selective Kernel Networks},
author={Xiang Li and Wenhai Wang and Xiaolin Hu and Jian Yang},
year={2019},
eprint={1903.06586},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
BENCHMARK | MODEL | METRIC NAME | METRIC VALUE | GLOBAL RANK |
---|---|---|---|---|
ImageNet | skresnet34 | Top 1 Accuracy | 76.93% | # 200 |
Top 5 Accuracy | 93.32% | # 200 | ||
ImageNet | skresnet18 | Top 1 Accuracy | 73.03% | # 253 |
Top 5 Accuracy | 91.17% | # 253 |