Xception

Chollet, 2017 — Xception: Deep Learning with Depthwise Separable Convolutions (arXiv:1610.02357)

Lucid port of timm/legacy_xception.tf_in1k, converted to Lucid-native safetensors.

Available weights

Tag	acc@1	acc@5	Params	GFLOPs	Size	Source
TF_IN1K (default)	79.0	—	22.9M	—	87.42 MB	timm

Usage

import lucid.models as models
from lucid.models.weights import XceptionWeights

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

# explicit tag (enum or string)
model = models.xception_cls(weights=XceptionWeights.TF_IN1K)
model = models.xception_cls(pretrained="TF_IN1K")

# preprocessing travels with the weights
weights = XceptionWeights.TF_IN1K
preprocess = weights.transforms()
logits = model(preprocess(image)[None]).logits

Conversion

Converted from timm/legacy_xception.tf_in1k via python -m tools.convert_weights xception --tag TF_IN1K. Key mapping + numerical parity verified against the source.

License

apache-2.0 — inherited from the original weights.

Citation

@inproceedings{chollet2017xception,
  title={Xception: Deep Learning with Depthwise Separable Convolutions},
  author={Chollet, Fran\c{c}ois},
  booktitle={CVPR}, year={2017}
}