I'm paying attention to Masked Auto Encoder (MAE) results lately since they appear to improve on the (speed/accuracy) performance of diffusion models. For example this SotA result released yesterday: https://t.co/rXaEJu5ePn Code: https://t.co/Ooa4HXQiaH
Contrastive Masked Autoencoders are Stronger Vision Learners 対照的なマスクされたオートエンコーダーは、より強力な視覚学習者です 2022-11-28T09:11:01+00:00 arXiv: https://t.co/hyxDwsq5hm 英/日サマリ↓ https://t.co/9vroXoyhNr
RT @_akhaliq: Contrastive Masked Autoencoders are Stronger Vision Learners abs: https://t.co/cexT7y5fxj CMAE-Base achieves 85.3% top-1 acc…
RT @_akhaliq: Contrastive Masked Autoencoders are Stronger Vision Learners abs: https://t.co/cexT7y5fxj CMAE-Base achieves 85.3% top-1 acc…