Saved in:
Bibliographic Details
Main Authors: Zhuo, Wei, Tang, Zhiyue, Xue, Wufeng, Ding, Hao, Ji, Junkai, Shen, Linlin
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.15669
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918210802024448
author Zhuo, Wei
Tang, Zhiyue
Xue, Wufeng
Ding, Hao
Ji, Junkai
Shen, Linlin
author_facet Zhuo, Wei
Tang, Zhiyue
Xue, Wufeng
Ding, Hao
Ji, Junkai
Shen, Linlin
contents Few-shot semantic segmentation has attracted growing interest for its ability to generalize to novel object categories using only a few annotated samples. To address data scarcity, recent methods incorporate multiple foundation models to improve feature transferability and segmentation performance. However, they often rely on dual-branch architectures that combine pre-trained encoders to leverage complementary strengths, a design that limits flexibility and efficiency. This raises a fundamental question: can we build a unified model that integrates knowledge from different foundation architectures? Achieving this is, however, challenging due to the misalignment between class-agnostic segmentation capabilities and fine-grained discriminative representations. To this end, we present UINO-FSS, a novel framework built on the key observation that early-stage DINOv2 features exhibit distribution consistency with SAM's output embeddings. This consistency enables the integration of both models' knowledge into a single-encoder architecture via coarse-to-fine multimodal distillation. In particular, our segmenter consists of three core components: a bottleneck adapter for embedding alignment, a meta-visual prompt generator that leverages dense similarity volumes and semantic embeddings, and a mask decoder. Using hierarchical cross-model distillation, we effectively transfer SAM's knowledge into the segmenter, further enhanced by Mamba-based 4D correlation mining on support-query pairs. Extensive experiments on PASCAL-5$^i$ and COCO-20$^i$ show that UINO-FSS achieves new state-of-the-art results under the 1-shot setting, with mIoU of 80.6 (+3.8%) on PASCAL-5$^i$ and 64.5 (+4.1%) on COCO-20$^i$, demonstrating the effectiveness of our unified approach.
format Preprint
id arxiv_https___arxiv_org_abs_2504_15669
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle UINO-FSS: Unifying Representation Learning and Few-shot Segmentation via Hierarchical Distillation and Mamba-HyperCorrelation
Zhuo, Wei
Tang, Zhiyue
Xue, Wufeng
Ding, Hao
Ji, Junkai
Shen, Linlin
Computer Vision and Pattern Recognition
Few-shot semantic segmentation has attracted growing interest for its ability to generalize to novel object categories using only a few annotated samples. To address data scarcity, recent methods incorporate multiple foundation models to improve feature transferability and segmentation performance. However, they often rely on dual-branch architectures that combine pre-trained encoders to leverage complementary strengths, a design that limits flexibility and efficiency. This raises a fundamental question: can we build a unified model that integrates knowledge from different foundation architectures? Achieving this is, however, challenging due to the misalignment between class-agnostic segmentation capabilities and fine-grained discriminative representations. To this end, we present UINO-FSS, a novel framework built on the key observation that early-stage DINOv2 features exhibit distribution consistency with SAM's output embeddings. This consistency enables the integration of both models' knowledge into a single-encoder architecture via coarse-to-fine multimodal distillation. In particular, our segmenter consists of three core components: a bottleneck adapter for embedding alignment, a meta-visual prompt generator that leverages dense similarity volumes and semantic embeddings, and a mask decoder. Using hierarchical cross-model distillation, we effectively transfer SAM's knowledge into the segmenter, further enhanced by Mamba-based 4D correlation mining on support-query pairs. Extensive experiments on PASCAL-5$^i$ and COCO-20$^i$ show that UINO-FSS achieves new state-of-the-art results under the 1-shot setting, with mIoU of 80.6 (+3.8%) on PASCAL-5$^i$ and 64.5 (+4.1%) on COCO-20$^i$, demonstrating the effectiveness of our unified approach.
title UINO-FSS: Unifying Representation Learning and Few-shot Segmentation via Hierarchical Distillation and Mamba-HyperCorrelation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2504.15669