Pretrained weights for lucid.models.inception_v4_cls.
Szegedy et al. AAAI 2017 (42.7 M params, top-1 80.14%).
Attributes
TF_IN1KWeightEntryTensorFlow-Slim ImageNet-1k checkpoint (top-1 80.144% / top-5
94.982% at 299, per timm's
results-imagenet.csv for this
exact tag), sourced from timm/inception_v4.tf_in1k. The
paper's Table 2 reports 80.0% / 95.0% single-crop for its own
evaluation.DEFAULTWeightEntryAlias for
TF_IN1K.Notes
Reference: Szegedy, Ioffe, Vanhoucke, Alemi, "Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning", AAAI 2017 (arXiv:1602.07261).
Examples
>>> from lucid.models.weights import InceptionV4Weights
>>> InceptionV4Weights.DEFAULT is InceptionV4Weights.TF_IN1K
True
>>> tf = InceptionV4Weights.TF_IN1K.transforms()
>>> tf.crop_size, tf.resize_size, tf.interpolation
(299, 341, 'bicubic')
>>> tuple(tf.mean), tuple(tf.std)
((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))