Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling

TL;DR

This paper introduces a novel unsupervised semantic segmentation method that leverages depth information to improve feature correlation and sampling, resulting in better performance on benchmark datasets.

Contribution

It integrates depth-guided feature correlation and 3D sampling techniques into unsupervised segmentation, advancing the state-of-the-art in structure-aware learning.

Findings

Significant performance improvements on multiple benchmarks.

Effective use of depth information for feature correlation.

Enhanced feature sampling with 3D techniques.

Abstract

Traditionally, training neural networks to perform semantic segmentation required expensive human-made annotations. But more recently, advances in the field of unsupervised learning have made significant progress on this issue and towards closing the gap to supervised algorithms. To achieve this, semantic knowledge is distilled by learning to correlate randomly sampled features from images across an entire dataset. In this work, we build upon these advances by incorporating information about the structure of the scene into the training process through the use of depth information. We achieve this by (1) learning depth-feature correlation by spatially correlate the feature maps with the depth maps to induce knowledge about the structure of the scene and (2) implementing farthest-point sampling to more effectively select relevant features by utilizing 3D sampling techniques on depth information of the scene. Finally, we demonstrate the effectiveness of our technical contributions through extensive experimentation and present significant improvements in performance across multiple benchmark datasets.