MaxViT and CoAtNet

Category: Hybrid · Code: timm/models/maxxvit.py · Architectures: 58 · Pretrained weights: 52 (43 with ImageNet-1k results)

From-scratch implementations of MaxViT (MBConv blocks with multi-axis attention, blocked local and dilated grid attention) and CoAtNet (convolution stages followed by relative attention stages), plus timm variants (_rw configs, MaxxViT).

Papers

• MaxViT: Multi-Axis Vision Transformer
• CoAtNet: Marrying Convolution and Attention for All Data Sizes

Usage

See the Quickstart for loading pretrained weights, image preprocessing and fine-tuning, and Feature Extraction for using the models as backbones.

>>> import timm
>>> model = timm.create_model('maxxvitv2_nano_rw_256.sw_in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)

Model API support

✓ supported, ✓ (square) works for square sizes only, ✗ not supported, – not needed (the model already accepts any input size). Checked by running each feature on a randomly initialized model (MaxxVit with maxvit_pico_rw_256 at 256), other architectures of the same class are expected to behave the same. See the Model API page for details.

Architectures

GMACs and parameter counts are for the default input size and classifier of each architecture, from the benchmark results.

Pretrained weights

Top-1 is ImageNet-1k validation accuracy from results-imagenet.csv at the listed eval size, - if the weight has no ImageNet-1k result (e.g. pretraining or image-text weights). Classes is the size of the pretrained classifier, none for weights without one.

Citation

@article{tu2022maxvit,
  title={MaxViT: Multi-Axis Vision Transformer},
  author={Tu, Zhengzhong and Talebi, Hossein and Zhang, Han and Yang, Feng and Milanfar, Peyman and Bovik, Alan and Li, Yinxiao},
  journal={ECCV},
  year={2022},
}

@article{DBLP:journals/corr/abs-2106-04803,
  author    = {Zihang Dai and Hanxiao Liu and Quoc V. Le and Mingxing Tan},
  title     = {CoAtNet: Marrying Convolution and Attention for All Data Sizes},
  journal   = {CoRR},
  volume    = {abs/2106.04803},
  year      = {2021}
}