Chamfer distance's data source

[threegold116]

in the paper “ VILA: On Pre-training for Visual Language Models” 's "The deep embedding alignment hypothesis." part , the Chamfer distance is interesting and useful. And I want to konw how it is calculate ? and what is the image source and the text source? Thank you very mcuh!

[tonylins]

Hi, thanks for your interests in our work!

An example code to measure the Shamfer distance (cosine) is:

x = torch.randn(32, 128)  # N, D
y = torch.randn(32, 128)  # N, D
x = x / torch.norm(x, 1, keepdim=True)
y = y / torch.norm(x, 1, keepdim=True)
sim = x @ y.T  # cosine similarity
dist = 0.5 * (sim.amax(1).mean() + sim.amax(0).mean())

We used a hold out set from the training mix to measure the distance.

[threegold116]

Thank you very much! I understand the process.