Pretrained weights for lucid.models.mobilenet_v4_conv_small_cls.
Qin et al. ECCV 2024 Conv-Small (3.8 M params, top-1 73.76%).
Attributes
E2400_R224_IN1KWeightEntryImageNet-1k checkpoint trained 2400 epochs at 224x224 (top-1
73.756% / top-5 91.430% at 224, per timm's
results-imagenet.csv for this exact tag), sourced from
timm/mobilenetv4_conv_small.e2400_r224_in1k.DEFAULTWeightEntryAlias for
E2400_R224_IN1K.Notes
Reference: Qin et al., "MobileNetV4: Universal Models for the Mobile Ecosystem", ECCV 2024 (arXiv:2404.10518).
Examples
>>> from lucid.models.weights import MobileNetV4ConvSmallWeights
>>> list(MobileNetV4ConvSmallWeights.__members__)
['E2400_R224_IN1K', 'DEFAULT']
>>> tf = MobileNetV4ConvSmallWeights.DEFAULT.transforms()
>>> tf.crop_size, tf.resize_size, tf.interpolation
(224, 256, 'bicubic')