lucid-dl/mobilenet-v2

MobileNet V2

Sandler et al., 2018 — MobileNetV2: Inverted Residuals and Linear Bottlenecks (arXiv:1801.04381)

Lucid port of torchvision/MobileNet_V2_Weights.IMAGENET1K_V1, converted to Lucid-native safetensors.

Available weights

Tag	acc@1	acc@5	Params	GFLOPs	Size	Source
IMAGENET1K_V1 (default)	71.878	90.286	3.5M	0.301	13.53 MB	torchvision

Usage

import lucid.models as models
from lucid.models.weights import MobileNetV2Weights

# default tag
model = models.mobilenet_v2_cls(pretrained=True)

# explicit tag (enum or string)
model = models.mobilenet_v2_cls(weights=MobileNetV2Weights.IMAGENET1K_V1)
model = models.mobilenet_v2_cls(pretrained="IMAGENET1K_V1")

# preprocessing travels with the weights
weights = MobileNetV2Weights.IMAGENET1K_V1
preprocess = weights.transforms()
logits = model(preprocess(image)[None]).logits

Conversion

Converted from torchvision/MobileNet_V2_Weights.IMAGENET1K_V1 via python -m tools.convert_weights mobilenet_v2 --tag IMAGENET1K_V1. Key mapping + numerical parity verified against the source.

License

bsd-3-clause — inherited from the original weights.

Citation

@inproceedings{sandler2018mobilenetv2,
  title={MobileNetV2: Inverted Residuals and Linear Bottlenecks},
  author={Sandler, Mark and Howard, Andrew and Zhu, Menglong and Zhmoginov, Andrey and Chen, Liang-Chieh},
  booktitle={CVPR}, year={2018}
}