SOTAVerified

Hierarchical Dense Correlation Distillation for Few-Shot Segmentation

2023-03-26CVPR 2023Code Available1· sign in to hype

Bohao Peng, Zhuotao Tian, Xiaoyang Wu, Chenyao Wang, Shu Liu, Jingyong Su, Jiaya Jia

Code Available — Be the first to reproduce this paper.

Reproduce

Code

Abstract

Few-shot semantic segmentation (FSS) aims to form class-agnostic models segmenting unseen classes with only a handful of annotations. Previous methods limited to the semantic feature and prototype representation suffer from coarse segmentation granularity and train-set overfitting. In this work, we design Hierarchically Decoupled Matching Network (HDMNet) mining pixel-level support correlation based on the transformer architecture. The self-attention modules are used to assist in establishing hierarchical dense features, as a means to accomplish the cascade matching between query and support features. Moreover, we propose a matching module to reduce train-set overfitting and introduce correlation distillation leveraging semantic correspondence from coarse resolution to boost fine-grained segmentation. Our method performs decently in experiments. We achieve 50.0\% mIoU on ~dataset one-shot setting and 56.0\% on five-shot segmentation, respectively.

Tasks

Benchmark Results

DatasetModelMetricClaimedVerifiedStatus
COCO-20i (1-shot)HDMNet (VGG-16)Mean IoU45.9Unverified
COCO-20i (1-shot)HDMNet (ResNet-50)Mean IoU50Unverified
COCO-20i (5-shot)HDMNet (ResNet-50)Mean IoU56Unverified
COCO-20i (5-shot)HDMNet (VGG-16)Mean IoU52.4Unverified
PASCAL-5i (1-Shot)HDMNet (VGG-16)Mean IoU65.1Unverified
PASCAL-5i (1-Shot)HDMNet (ResNet-50)Mean IoU69.4Unverified
PASCAL-5i (5-Shot)HDMNet (ResNet-50)Mean IoU71.8Unverified
PASCAL-5i (5-Shot)HDMNet (VGG-16)Mean IoU69.3Unverified

Reproductions