Self-Supervised Image Representation Learning with Geometric Set Consistency

TL;DR

This paper introduces a self-supervised learning method that leverages 3D geometric consistency priors to improve 2D image representations, leading to better performance on various downstream tasks.

Contribution

It integrates 3D geometric consistency into contrastive learning, enhancing feature learning without semantic labels, and demonstrates superior results on real-world indoor scene datasets.

Findings

Improved semantic segmentation accuracy

Enhanced object detection performance

Superior instance segmentation results

Abstract

We propose a method for self-supervised image representation learning under the guidance of 3D geometric consistency. Our intuition is that 3D geometric consistency priors such as smooth regions and surface discontinuities may imply consistent semantics or object boundaries, and can act as strong cues to guide the learning of 2D image representations without semantic labels. Specifically, we introduce 3D geometric consistency into a contrastive learning framework to enforce the feature consistency within image views. We propose to use geometric consistency sets as constraints and adapt the InfoNCE loss accordingly. We show that our learned image representations are general. By fine-tuning our pre-trained representations for various 2D image-based downstream tasks, including semantic segmentation, object detection, and instance segmentation on real-world indoor scene datasets, we achieve superior performance compared with state-of-the-art methods.