Open-world Semantic Segmentation via Contrasting and Clustering Vision-Language Embedding

TL;DR

This paper introduces ViL-Seg, a novel open-world semantic segmentation approach that leverages image-caption data and contrastive learning to segment arbitrary categories without dense annotations.

Contribution

It pioneers a segmentation pipeline that learns from image-caption data alone, eliminating the need for dense annotations and enabling recognition of arbitrary open-world categories.

Findings

Outperforms existing zero-shot segmentation methods on three benchmarks.

Successfully segments objects of arbitrary categories without dense annotations.

Uses contrastive learning and online clustering to enhance segmentation quality.

Abstract

To bridge the gap between supervised semantic segmentation and real-world applications that acquires one model to recognize arbitrary new concepts, recent zero-shot segmentation attracts a lot of attention by exploring the relationships between unseen and seen object categories, yet requiring large amounts of densely-annotated data with diverse base classes. In this paper, we propose a new open-world semantic segmentation pipeline that makes the first attempt to learn to segment semantic objects of various open-world categories without any efforts on dense annotations, by purely exploiting the image-caption data that naturally exist on the Internet. Our method, Vision-language-driven Semantic Segmentation (ViL-Seg), employs an image and a text encoder to generate visual and text embeddings for the image-caption data, with two core components that endow its segmentation ability: First, the image encoder is jointly trained with a vision-based contrasting and a cross-modal contrasting, which encourage the visual embeddings to preserve both fine-grained semantics and high-level category information that are crucial for the segmentation task. Furthermore, an online clustering head is devised over the image encoder, which allows to dynamically segment the visual embeddings into distinct semantic groups such that they can be classified by comparing with various text embeddings to complete our segmentation pipeline. Experiments show that without using any data with dense annotations, our method can directly segment objects of arbitrary categories, outperforming zero-shot segmentation methods that require data labeling on three benchmark datasets.