CrossViT-9 backbone — embed_dims=(128, 256), depths
((1, 3, 0))×3, 4 heads. ~8.6M params (paper Table 2).
Model Size
Examples
>>> import lucid
>>> from lucid.models.vision.crossvit import crossvit_9
>>> model = crossvit_9()
>>> out = model(lucid.randn(1, 3, 240, 240))
>>> out.last_hidden_state.shape
(1, 384)
CrossViT works at 240 pixels rather than 224: the two branches
use 12- and 16-pixel patches, and 240 is what both divide.