Unsupervised Multi-object Segmentation Using Attention and Soft-argmax

TL;DR

This paper presents an unsupervised architecture for multi-object segmentation that leverages attention mechanisms and a transformer encoder to improve detection and segmentation accuracy in complex scenes.

Contribution

The novel architecture combines attention, transformer, and autoencoder components for unsupervised multi-object segmentation, outperforming previous methods on synthetic benchmarks.

Findings

Significantly outperforms state-of-the-art on synthetic benchmarks

Uses attention and transformer for occlusion handling

Effective background reconstruction with autoencoder

Abstract

We introduce a new architecture for unsupervised object-centric representation learning and multi-object detection and segmentation, which uses a translation-equivariant attention mechanism to predict the coordinates of the objects present in the scene and to associate a feature vector to each object. A transformer encoder handles occlusions and redundant detections, and a convolutional autoencoder is in charge of background reconstruction. We show that this architecture significantly outperforms the state of the art on complex synthetic benchmarks.