Pretrained weights for lucid.models.mobilenet_v4_conv_medium_cls.
Qin et al. ECCV 2024 Conv-Medium (9.7 M params, top-1 79.92%).
Attributes
E500_R256_IN1KWeightEntryImageNet-1k checkpoint trained 500 epochs at 256x256 (top-1
79.916% / top-5 95.188% at 256, per timm's
results-imagenet.csv for this exact tag), sourced from
timm/mobilenetv4_conv_medium.e500_r256_in1k.DEFAULTWeightEntryAlias for
E500_R256_IN1K.Notes
Reference: Qin et al., "MobileNetV4: Universal Models for the Mobile Ecosystem", ECCV 2024 (arXiv:2404.10518).
Examples
>>> from lucid.models.weights import MobileNetV4ConvMediumWeights
>>> MobileNetV4ConvMediumWeights.DEFAULT is (
... MobileNetV4ConvMediumWeights.E500_R256_IN1K
... )
True
>>> tf = MobileNetV4ConvMediumWeights.DEFAULT.transforms()
>>> tf.crop_size, tf.resize_size
(256, 269)