FrozenSeg: Harmonizing Frozen Foundation Models for Open-Vocabulary Segmentation

TL;DR

FrozenSeg effectively combines spatial and semantic knowledge from pre-trained foundation models like SAM and CLIP to improve open-vocabulary segmentation, achieving state-of-the-art zero-shot performance with minimal training overhead.

Contribution

The paper introduces FrozenSeg, a novel framework that integrates frozen spatial and semantic foundation models for improved open-vocabulary segmentation.

Findings

Achieves state-of-the-art zero-shot segmentation results.

Utilizes a lightweight transformer decoder trained on COCO data.

Demonstrates significant performance gains across multiple benchmarks.

Abstract

Open-vocabulary segmentation poses significant challenges, as it requires segmenting and recognizing objects across an open set of categories in unconstrained environments. Building on the success of powerful vision-language (ViL) foundation models, such as CLIP, recent efforts sought to harness their zero-short capabilities to recognize unseen categories. Despite notable performance improvements, these models still encounter the critical issue of generating precise mask proposals for unseen categories and scenarios, resulting in inferior segmentation performance eventually. To address this challenge, we introduce a novel approach, FrozenSeg, designed to integrate spatial knowledge from a localization foundation model (e.g., SAM) and semantic knowledge extracted from a ViL model (e.g., CLIP), in a synergistic framework. Taking the ViL model's visual encoder as the feature backbone, we inject the space-aware feature into the learnable queries and CLIP features within the transformer decoder. In addition, we devise a mask proposal ensemble strategy for further improving the recall rate and mask quality. To fully exploit pre-trained knowledge while minimizing training overhead, we freeze both foundation models, focusing optimization efforts solely on a lightweight transformer decoder for mask proposal generation-the performance bottleneck. Extensive experiments demonstrate that FrozenSeg advances state-of-the-art results across various segmentation benchmarks, trained exclusively on COCO panoptic data, and tested in a zero-shot manner. Code is available at https://github.com/chenxi52/FrozenSeg.