Architecture | Batch Normalization, Dense Connections, Dropout, Global Average Pooling, Grouped Convolution, MixConv, Squeeze-and-Excitation Block, Swish |
---|---|
ID | mixnet_l |
SHOW MORE |
Architecture | Batch Normalization, Dense Connections, Dropout, Global Average Pooling, Grouped Convolution, MixConv, Squeeze-and-Excitation Block, Swish |
---|---|
ID | mixnet_m |
SHOW MORE |
Architecture | Batch Normalization, Dense Connections, Dropout, Global Average Pooling, Grouped Convolution, MixConv, Squeeze-and-Excitation Block, Swish |
---|---|
ID | mixnet_s |
SHOW MORE |
Architecture | Batch Normalization, Dense Connections, Dropout, Global Average Pooling, Grouped Convolution, MixConv, Squeeze-and-Excitation Block, Swish |
---|---|
ID | mixnet_xl |
SHOW MORE |
MixNet is a type of convolutional neural network discovered via AutoML that utilises MixConvs instead of regular depthwise convolutions.
To load a pretrained model:
import timm
m = timm.create_model('mixnet_s', pretrained=True)
m.eval()
Replace the model name with the variant you want to use, e.g. mixnet_s
. 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{tan2019mixconv,
title={MixConv: Mixed Depthwise Convolutional Kernels},
author={Mingxing Tan and Quoc V. Le},
year={2019},
eprint={1907.09595},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
MODEL | TOP 1 ACCURACY | TOP 5 ACCURACY |
---|---|---|
mixnet_xl | 80.47% | 94.93% |
mixnet_l | 78.98% | 94.18% |
mixnet_m | 77.27% | 93.42% |
mixnet_s | 75.99% | 92.79% |