Tuning-Free Universally-Supervised Semantic Segmentation

This work presents a tuning-free semantic segmentation framework based on classifying SAM masks, which is universally applicable to various types of supervision. Initially, we utilize CLIP’s zero-shot classification ability to generate pseudo-labels or perform open-vocabulary semantic seg...

Full description

Saved in:
Bibliographic Details
Main Authors: Xiaobo Yang, Xiaojin Gong
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10779462/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This work presents a tuning-free semantic segmentation framework based on classifying SAM masks, which is universally applicable to various types of supervision. Initially, we utilize CLIP’s zero-shot classification ability to generate pseudo-labels or perform open-vocabulary semantic segmentation. However, the misalignment between mask and CLIP text embeddings leads to suboptimal results. To address this issue, we propose discrimination-bias aligned CLIP to closely align mask and text embedding, offering an overhead-free performance gain. We then construct a global-local consistent classifier to classify SAM masks, which reveals the intrinsic structure of high-quality embeddings produced by DBA-CLIP and demonstrates robustness against noisy pseudo-labels. Extensive experiments validate the efficiency and effectiveness of our method, and we achieve state-of-the-art (SOTA) or competitive performance across various datasets and supervision types. Our code will be released upon acceptance.
ISSN:2169-3536