Unlocking Slot Attention by Changing Optimal Transport Costs
| Properties | |
|---|---|
| authors | Yan Zhang, David W. Zhang, Simon Lacoste-Julien, Cees G. M. Snoek, Gertjan J. Burghouts |
| year | 2023 |
| url | http://arxiv.org/abs/2301.13197 |
Abstract
Slot attention is a powerful method for objectcentric modeling in images and videos. However, its set-equivariance limits its ability to handle videos with a dynamic number of objects because it cannot break ties. To overcome this limitation, we first establish a connection between slot attention and optimal transport. Based on this new perspective we propose MESH (Minimize Entropy of Sinkhorn): a cross-attention module that combines the tiebreaking properties of unregularized optimal transport with the speed of regularized optimal transport. We evaluate slot attention using MESH on multiple object-centric learning benchmarks and find significant improvements over slot attention in every setting.