The Missing Point in Vision Transformers for Universal Image Segmentation

TL;DR

This paper introduces ViT-P, a two-stage vision transformer-based framework for universal image segmentation that decouples mask proposal and classification, improving accuracy and reducing annotation costs.

Contribution

The paper proposes ViT-P, a novel two-stage segmentation framework that integrates pre-trained vision transformers without modification, and leverages coarse annotations to enhance classification.

Findings

Achieves state-of-the-art PQ of 54.0 on ADE20K

Attains 87.4 mIoU on Cityscapes

Demonstrates effective use of coarse annotations

Abstract

Image segmentation remains a challenging task in computer vision, demanding robust mask generation and precise classification. Recent mask-based approaches yield high-quality masks by capturing global context. However, accurately classifying these masks, especially in the presence of ambiguous boundaries and imbalanced class distributions, remains an open challenge. In this work, we introduce ViT-P, a novel two-stage segmentation framework that decouples mask generation from classification. The first stage employs a proposal generator to produce class-agnostic mask proposals, while the second stage utilizes a point-based classification model built on the Vision Transformer (ViT) to refine predictions by focusing on mask central points. ViT-P serves as a pre-training-free adapter, allowing the integration of various pre-trained vision transformers without modifying their architecture, ensuring adaptability to dense prediction tasks. Furthermore, we demonstrate that coarse and bounding box annotations can effectively enhance classification without requiring additional training on fine annotation datasets, reducing annotation costs while maintaining strong performance. Extensive experiments across COCO, ADE20K, and Cityscapes datasets validate the effectiveness of ViT-P, achieving state-of-the-art results with 54.0 PQ on ADE20K panoptic segmentation, 87.4 mIoU on Cityscapes semantic segmentation, and 63.6 mIoU on ADE20K semantic segmentation. The code and pretrained models are available at: https://github.com/sajjad-sh33/ViT-P}{https://github.com/sajjad-sh33/ViT-P.