All Things ViTs || CVPR 2023 Tutorial || Hila Chefer and Sayak Paul

[@vi5hnupradeep]

Thanks for sharing this Sayak Paul . As always , amazing work in covering everything in so much detail.

[@lorenzoleongutierrez7927]

Thanks for sharing!

[@НиколайНовичков-е1э]

Thanks for sharing! It was very interesting!

[@tydsuper3122]

what a good presentation！

[@aritraroygosthipaty3662]

Amazing! Congratulations.

[@ritwikraha]

This is fantastic! Congratulations Sayak da! 🎉

[@sbeg-wv7fz]

Thanks for sharing

[@deepaksingh-vt2gq]

I have a doubt, at 19:41 (From Self-Attention to Cross-Attention) slide, at the bottom, shouldn't we group Q and K for Text and V for Image?

[@hoangminhnguyen435]

Cross Attention means you want to find the relevance between different objects inputs, so basically Q and K must come from different source, and K and V must come from same source because K and V are representation of same object with different scope.

[@amitpareek4215]

Congrats sir

[@soumyasarkar4100]

congrats !

[@reloto5665]

Is there a small mistake at 53:09?
Shouldn't it be 0.2+0.05+0.04 = 0.31 instead of 0.11?

[@SphereofTime]

1:43:16

[@mikhailandreev1595]

Sayak is the backbone of half the global ML community