Saved in:
Bibliographic Details
Main Authors: Yang, Xiaobo, Gong, Xiaojin
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.14294
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914807660150784
author Yang, Xiaobo
Gong, Xiaojin
author_facet Yang, Xiaobo
Gong, Xiaojin
contents This work presents a tuning-free semantic segmentation framework based on classifying SAM masks by CLIP, 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 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.
format Preprint
id arxiv_https___arxiv_org_abs_2405_14294
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Tuning-free Universally-Supervised Semantic Segmentation
Yang, Xiaobo
Gong, Xiaojin
Computer Vision and Pattern Recognition
This work presents a tuning-free semantic segmentation framework based on classifying SAM masks by CLIP, 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 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.
title Tuning-free Universally-Supervised Semantic Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2405.14294