Exploring the Role of the Bottleneck in Slot-Based Models Through Covariance Regularization

TL;DR

This paper investigates how constraining the bottleneck in slot-based models with covariance regularization affects their performance, aiming to improve image reconstruction and mask quality on real-world datasets.

Contribution

It introduces a loss-based method to regularize the bottleneck in slot-based models, enabling larger encoders and improving over baseline Slot Attention models.

Findings

Improved performance over baseline Slot Attention

Feature reconstruction outperforms image reconstruction

Bottleneck regularization enhances model capacity

Abstract

In this project we attempt to make slot-based models with an image reconstruction objective competitive with those that use a feature reconstruction objective on real world datasets. We propose a loss-based approach to constricting the bottleneck of slot-based models, allowing larger-capacity encoder networks to be used with Slot Attention without producing degenerate stripe-shaped masks. We find that our proposed method offers an improvement over the baseline Slot Attention model but does not reach the performance of \dinosaur on the COCO2017 dataset. Throughout this project, we confirm the superiority of a feature reconstruction objective over an image reconstruction objective and explore the role of the architectural bottleneck in slot-based models.