Title

HMFS: Hybrid Masking for Few-Shot Segmentation

Document Type

Article

Publication Title

arXiv

Abstract

We study few-shot semantic segmentation that aims to segment a target object from a query image when provided with a few annotated support images of the target class. Several recent methods resort to a feature masking (FM) technique, introduced by [1], to discard irrelevant feature activations to facilitate reliable segmentation mask prediction. A fundamental limitation of FM is the inability to preserve the fine-grained spatial details that affect the accuracy of segmentation mask, especially for small target objects. In this paper, we develop a simple, effective, and efficient approach to enhance feature masking (FM). We dub the enhanced FM as hybrid masking (HM). Specifically, we compensate for the loss of fine-grained spatial details in FM technique by investigating and leveraging a complementary basic input masking method [2]. To validate the effectiveness of HM, we instantiate it into a strong baseline [3], and coin the resulting framework as HMFS. Experimental results on three publicly available benchmarks reveal that HMFS outperforms the current state-of-the-art methods by visible margins.

MSC Codes 68T45 Copyright © 2022, The Authors. All rights reserved.

DOI

10.48550/arXiv.2203.12826

Publication Date

3-24-2022

Keywords

Frequency modulation, Semantics, Feature masking, Few-shot learning, Few-shot segmentation, Fine grained, Masking technique, Query images, Segmentation masks, Semantic segmentation, Shot segmentation, Target object, Computer Vision and Pattern Recognition (cs.CV)

Comments

IR Deposit conditions: non-described

Preprint available on arXiv

Link to Full Text

Share

COinS