lucid-dl/crossvit-base

CrossViT-B

Chen et al., 2021 — CrossViT: Cross-Attention Multi-Scale Vision Transformer for Image Classification (arXiv:2103.14899)

Lucid port of timm/crossvit_base_240.in1k, converted to Lucid-native safetensors.

Available weights

Tag	acc@1	acc@5	Params	GFLOPs	Size	Source
IN1K (default)	82.2	—	105.0M	—	400.67 MB	timm

Usage

import lucid.models as models
from lucid.models.weights import CrossViTBaseWeights

# default tag
model = models.crossvit_base_cls(pretrained=True)

# explicit tag (enum or string)
model = models.crossvit_base_cls(weights=CrossViTBaseWeights.IN1K)
model = models.crossvit_base_cls(pretrained="IN1K")

# preprocessing travels with the weights
weights = CrossViTBaseWeights.IN1K
preprocess = weights.transforms()
logits = model(preprocess(image)[None]).logits

Conversion

Converted from timm/crossvit_base_240.in1k via python -m tools.convert_weights crossvit_base --tag IN1K. Key mapping + numerical parity verified against the source.

License

apache-2.0 — inherited from the original weights.

Citation

@inproceedings{chen2021crossvit,
  title={CrossViT: Cross-Attention Multi-Scale Vision Transformer for Image Classification},
  author={Chen, Chun-Fu (Richard) and Fan, Quanfu and Panda, Rameswar},
  booktitle={ICCV}, year={2021}
}