Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Hypercorrelation Squeeze for Few-Shot Segmentation

About

Few-shot semantic segmentation aims at learning to segment a target object from a query image using only a few annotated support images of the target class. This challenging task requires to understand diverse levels of visual cues and analyze fine-grained correspondence relations between the query and the support images. To address the problem, we propose Hypercorrelation Squeeze Networks (HSNet) that leverages multi-level feature correlation and efficient 4D convolutions. It extracts diverse features from different levels of intermediate convolutional layers and constructs a collection of 4D correlation tensors, i.e., hypercorrelations. Using efficient center-pivot 4D convolutions in a pyramidal architecture, the method gradually squeezes high-level semantic and low-level geometric cues of the hypercorrelation into precise segmentation masks in coarse-to-fine manner. The significant performance improvements on standard few-shot segmentation benchmarks of PASCAL-5i, COCO-20i, and FSS-1000 verify the efficacy of the proposed method.

Juhong Min, Dahyun Kang, Minsu Cho• 2021

Related benchmarks

TaskDatasetResultRank
Few-shot SegmentationPASCAL-5i
mIoU (Fold 0)72.2
325
Few-shot Semantic SegmentationPASCAL-5^i (test)
FB-IoU80.6
177
Few-shot SegmentationCOCO 20^i (test)
mIoU49.5
174
Semantic segmentationCOCO-20i
mIoU (Mean)49.5
132
Few-shot Semantic SegmentationCOCO-20i
mIoU55.1
115
Semantic segmentationPASCAL-5i
Mean mIoU73.8
111
Semantic segmentationPASCAL-5^i (test)--
107
Semantic segmentationPASCAL 5-shot 5i
Mean mIoU70.4
100
Few-shot Semantic SegmentationPASCAL-5i
mIoU70.4
96
Few-shot Semantic SegmentationCOCO 5-shot 20i
mIoU49.5
85
Showing 10 of 117 rows
...

Other info

Code

Follow for update