TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/HM: Hybrid Masking for Few-Shot Segmentation

HM: Hybrid Masking for Few-Shot Segmentation

Seonghyeon Moon, Samuel S. Sohn, Honglu Zhou, Sejong Yoon, Vladimir Pavlovic, Muhammad Haris Khan, Mubbasir Kapadia

2022-03-24SegmentationFew-Shot Semantic SegmentationSemantic Segmentation
PaperPDFCode(official)

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 to discard irrelevant feature activations which eventually facilitates the reliable prediction of segmentation mask. 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. Experiments have been conducted on three publicly available benchmarks with strong few-shot segmentation (FSS) baselines. We empirically show improved performance against the current state-of-the-art methods by visible margins across different benchmarks. Our code and trained models are available at: https://github.com/moonsh/HM-Hybrid-Masking

Results

TaskDatasetMetricValueModel
Few-Shot LearningFSS-1000 (5-shot)Mean IoU90.5VAT (HM, ResNet-101)
Few-Shot LearningFSS-1000 (5-shot)Mean IoU89.9VAT (HM, ResNet-50)
Few-Shot LearningFSS-1000 (5-shot)Mean IoU88.5HSNet (HM, ResNet-101)
Few-Shot LearningFSS-1000 (5-shot)Mean IoU88HSNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (5-shot)FB-IoU73.3ASNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i (5-shot)Mean IoU50.6ASNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i (5-shot)FB-IoU72.9HSNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i (5-shot)Mean IoU50.6HSNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i (5-shot)FB-IoU72.2HSNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (5-shot)Mean IoU49.4HSNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (5-shot)FB-IoU72.2ASNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (5-shot)Mean IoU48.4ASNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (5-shot)FB-IoU71.8VAT (HM, ResNet-50)
Few-Shot LearningCOCO-20i (5-shot)Mean IoU48.3VAT (HM, ResNet-50)
Few-Shot LearningCOCO-20i -> Pascal VOC (1-shot)Mean IoU66.5HSNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i -> Pascal VOC (1-shot)Mean IoU65.2HSNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i -> Pascal VOC (1-shot)Mean IoU65.1VAT (HM, ResNet-50)
Few-Shot LearningFSS-1000 (1-shot)Mean IoU90.2VAT (HM, ResNet-101)
Few-Shot LearningFSS-1000 (1-shot)Mean IoU89.4VAT (HM, ResNet-50)
Few-Shot LearningFSS-1000 (1-shot)Mean IoU87.8HSNet (HM, ResNet-101)
Few-Shot LearningFSS-1000 (1-shot)Mean IoU87.1HSNet (HM, ResNet-50)
Few-Shot LearningPASCAL-5i (1-Shot)FB-IoU79.4VAT (HM, ResNet-101)
Few-Shot LearningPASCAL-5i (1-Shot)Mean IoU67.8VAT (HM, ResNet-101)
Few-Shot LearningPASCAL-5i (1-Shot)FB-IoU77.8HSNet (HM, ResNet-101)
Few-Shot LearningPASCAL-5i (1-Shot)Mean IoU66.7HSNet (HM, ResNet-101)
Few-Shot LearningPASCAL-5i (1-Shot)FB-IoU77.1VAT (HM, ResNet-50)
Few-Shot LearningPASCAL-5i (1-Shot)Mean IoU65.8VAT (HM, ResNet-50)
Few-Shot LearningPASCAL-5i (1-Shot)FB-IoU76.5HSNet (HM, ResNet-50)
Few-Shot LearningPASCAL-5i (1-Shot)Mean IoU65HSNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (1-shot)FB-IoU71.5HSNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i (1-shot)Mean IoU46.5HSNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i (1-shot)FB-IoU71.1ASNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i (1-shot)Mean IoU45.9ASNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i (1-shot)FB-IoU70.4ASNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (1-shot)Mean IoU44.7ASNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (1-shot)FB-IoU70.8HSNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (1-shot)Mean IoU44.3HSNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i (1-shot)FB-IoU70VAT (HM, ResNet-50)
Few-Shot LearningCOCO-20i (1-shot)Mean IoU43.2VAT (HM, ResNet-50)
Few-Shot LearningPASCAL-5i (5-Shot)FB-IoU81.5VAT (HM, ResNet-101)
Few-Shot LearningPASCAL-5i (5-Shot)Mean IoU70.9VAT (HM, ResNet-101)
Few-Shot LearningPASCAL-5i (5-Shot)FB-IoU79.7HSNet (HM, ResNet-101)
Few-Shot LearningPASCAL-5i (5-Shot)Mean IoU69.3HSNet (HM, ResNet-101)
Few-Shot LearningPASCAL-5i (5-Shot)FB-IoU78.5VAT (HM, ResNet-50)
Few-Shot LearningPASCAL-5i (5-Shot)Mean IoU68.2VAT (HM, ResNet-50)
Few-Shot LearningPASCAL-5i (5-Shot)FB-IoU77.7HSNet (HM, ResNet-50)
Few-Shot LearningPASCAL-5i (5-Shot)Mean IoU67.1HSNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i -> Pascal VOC (5-shot)Mean IoU70.9HSNet (HM, ResNet-101)
Few-Shot LearningCOCO-20i -> Pascal VOC (5-shot)Mean IoU69.7HSNet (HM, ResNet-50)
Few-Shot LearningCOCO-20i -> Pascal VOC (5-shot)Mean IoU69.7VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationFSS-1000 (5-shot)Mean IoU90.5VAT (HM, ResNet-101)
Few-Shot Semantic SegmentationFSS-1000 (5-shot)Mean IoU89.9VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationFSS-1000 (5-shot)Mean IoU88.5HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationFSS-1000 (5-shot)Mean IoU88HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)FB-IoU73.3ASNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)Mean IoU50.6ASNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)FB-IoU72.9HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)Mean IoU50.6HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)FB-IoU72.2HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)Mean IoU49.4HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)FB-IoU72.2ASNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)Mean IoU48.4ASNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)FB-IoU71.8VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (5-shot)Mean IoU48.3VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i -> Pascal VOC (1-shot)Mean IoU66.5HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i -> Pascal VOC (1-shot)Mean IoU65.2HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i -> Pascal VOC (1-shot)Mean IoU65.1VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationFSS-1000 (1-shot)Mean IoU90.2VAT (HM, ResNet-101)
Few-Shot Semantic SegmentationFSS-1000 (1-shot)Mean IoU89.4VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationFSS-1000 (1-shot)Mean IoU87.8HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationFSS-1000 (1-shot)Mean IoU87.1HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationPASCAL-5i (1-Shot)FB-IoU79.4VAT (HM, ResNet-101)
Few-Shot Semantic SegmentationPASCAL-5i (1-Shot)Mean IoU67.8VAT (HM, ResNet-101)
Few-Shot Semantic SegmentationPASCAL-5i (1-Shot)FB-IoU77.8HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationPASCAL-5i (1-Shot)Mean IoU66.7HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationPASCAL-5i (1-Shot)FB-IoU77.1VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationPASCAL-5i (1-Shot)Mean IoU65.8VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationPASCAL-5i (1-Shot)FB-IoU76.5HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationPASCAL-5i (1-Shot)Mean IoU65HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)FB-IoU71.5HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)Mean IoU46.5HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)FB-IoU71.1ASNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)Mean IoU45.9ASNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)FB-IoU70.4ASNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)Mean IoU44.7ASNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)FB-IoU70.8HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)Mean IoU44.3HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)FB-IoU70VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i (1-shot)Mean IoU43.2VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationPASCAL-5i (5-Shot)FB-IoU81.5VAT (HM, ResNet-101)
Few-Shot Semantic SegmentationPASCAL-5i (5-Shot)Mean IoU70.9VAT (HM, ResNet-101)
Few-Shot Semantic SegmentationPASCAL-5i (5-Shot)FB-IoU79.7HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationPASCAL-5i (5-Shot)Mean IoU69.3HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationPASCAL-5i (5-Shot)FB-IoU78.5VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationPASCAL-5i (5-Shot)Mean IoU68.2VAT (HM, ResNet-50)
Few-Shot Semantic SegmentationPASCAL-5i (5-Shot)FB-IoU77.7HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationPASCAL-5i (5-Shot)Mean IoU67.1HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i -> Pascal VOC (5-shot)Mean IoU70.9HSNet (HM, ResNet-101)
Few-Shot Semantic SegmentationCOCO-20i -> Pascal VOC (5-shot)Mean IoU69.7HSNet (HM, ResNet-50)
Few-Shot Semantic SegmentationCOCO-20i -> Pascal VOC (5-shot)Mean IoU69.7VAT (HM, ResNet-50)
Meta-LearningFSS-1000 (5-shot)Mean IoU90.5VAT (HM, ResNet-101)
Meta-LearningFSS-1000 (5-shot)Mean IoU89.9VAT (HM, ResNet-50)
Meta-LearningFSS-1000 (5-shot)Mean IoU88.5HSNet (HM, ResNet-101)
Meta-LearningFSS-1000 (5-shot)Mean IoU88HSNet (HM, ResNet-50)
Meta-LearningCOCO-20i (5-shot)FB-IoU73.3ASNet (HM, ResNet-101)
Meta-LearningCOCO-20i (5-shot)Mean IoU50.6ASNet (HM, ResNet-101)
Meta-LearningCOCO-20i (5-shot)FB-IoU72.9HSNet (HM, ResNet-101)
Meta-LearningCOCO-20i (5-shot)Mean IoU50.6HSNet (HM, ResNet-101)
Meta-LearningCOCO-20i (5-shot)FB-IoU72.2HSNet (HM, ResNet-50)
Meta-LearningCOCO-20i (5-shot)Mean IoU49.4HSNet (HM, ResNet-50)
Meta-LearningCOCO-20i (5-shot)FB-IoU72.2ASNet (HM, ResNet-50)
Meta-LearningCOCO-20i (5-shot)Mean IoU48.4ASNet (HM, ResNet-50)
Meta-LearningCOCO-20i (5-shot)FB-IoU71.8VAT (HM, ResNet-50)
Meta-LearningCOCO-20i (5-shot)Mean IoU48.3VAT (HM, ResNet-50)
Meta-LearningCOCO-20i -> Pascal VOC (1-shot)Mean IoU66.5HSNet (HM, ResNet-101)
Meta-LearningCOCO-20i -> Pascal VOC (1-shot)Mean IoU65.2HSNet (HM, ResNet-50)
Meta-LearningCOCO-20i -> Pascal VOC (1-shot)Mean IoU65.1VAT (HM, ResNet-50)
Meta-LearningFSS-1000 (1-shot)Mean IoU90.2VAT (HM, ResNet-101)
Meta-LearningFSS-1000 (1-shot)Mean IoU89.4VAT (HM, ResNet-50)
Meta-LearningFSS-1000 (1-shot)Mean IoU87.8HSNet (HM, ResNet-101)
Meta-LearningFSS-1000 (1-shot)Mean IoU87.1HSNet (HM, ResNet-50)
Meta-LearningPASCAL-5i (1-Shot)FB-IoU79.4VAT (HM, ResNet-101)
Meta-LearningPASCAL-5i (1-Shot)Mean IoU67.8VAT (HM, ResNet-101)
Meta-LearningPASCAL-5i (1-Shot)FB-IoU77.8HSNet (HM, ResNet-101)
Meta-LearningPASCAL-5i (1-Shot)Mean IoU66.7HSNet (HM, ResNet-101)
Meta-LearningPASCAL-5i (1-Shot)FB-IoU77.1VAT (HM, ResNet-50)
Meta-LearningPASCAL-5i (1-Shot)Mean IoU65.8VAT (HM, ResNet-50)
Meta-LearningPASCAL-5i (1-Shot)FB-IoU76.5HSNet (HM, ResNet-50)
Meta-LearningPASCAL-5i (1-Shot)Mean IoU65HSNet (HM, ResNet-50)
Meta-LearningCOCO-20i (1-shot)FB-IoU71.5HSNet (HM, ResNet-101)
Meta-LearningCOCO-20i (1-shot)Mean IoU46.5HSNet (HM, ResNet-101)
Meta-LearningCOCO-20i (1-shot)FB-IoU71.1ASNet (HM, ResNet-101)
Meta-LearningCOCO-20i (1-shot)Mean IoU45.9ASNet (HM, ResNet-101)
Meta-LearningCOCO-20i (1-shot)FB-IoU70.4ASNet (HM, ResNet-50)
Meta-LearningCOCO-20i (1-shot)Mean IoU44.7ASNet (HM, ResNet-50)
Meta-LearningCOCO-20i (1-shot)FB-IoU70.8HSNet (HM, ResNet-50)
Meta-LearningCOCO-20i (1-shot)Mean IoU44.3HSNet (HM, ResNet-50)
Meta-LearningCOCO-20i (1-shot)FB-IoU70VAT (HM, ResNet-50)
Meta-LearningCOCO-20i (1-shot)Mean IoU43.2VAT (HM, ResNet-50)
Meta-LearningPASCAL-5i (5-Shot)FB-IoU81.5VAT (HM, ResNet-101)
Meta-LearningPASCAL-5i (5-Shot)Mean IoU70.9VAT (HM, ResNet-101)
Meta-LearningPASCAL-5i (5-Shot)FB-IoU79.7HSNet (HM, ResNet-101)
Meta-LearningPASCAL-5i (5-Shot)Mean IoU69.3HSNet (HM, ResNet-101)
Meta-LearningPASCAL-5i (5-Shot)FB-IoU78.5VAT (HM, ResNet-50)
Meta-LearningPASCAL-5i (5-Shot)Mean IoU68.2VAT (HM, ResNet-50)
Meta-LearningPASCAL-5i (5-Shot)FB-IoU77.7HSNet (HM, ResNet-50)
Meta-LearningPASCAL-5i (5-Shot)Mean IoU67.1HSNet (HM, ResNet-50)
Meta-LearningCOCO-20i -> Pascal VOC (5-shot)Mean IoU70.9HSNet (HM, ResNet-101)
Meta-LearningCOCO-20i -> Pascal VOC (5-shot)Mean IoU69.7HSNet (HM, ResNet-50)
Meta-LearningCOCO-20i -> Pascal VOC (5-shot)Mean IoU69.7VAT (HM, ResNet-50)

Related Papers

SeC: Advancing Complex Video Object Segmentation via Progressive Concept Construction2025-07-21Deep Learning-Based Fetal Lung Segmentation from Diffusion-weighted MRI Images and Lung Maturity Evaluation for Fetal Growth Restriction2025-07-17DiffOSeg: Omni Medical Image Segmentation via Multi-Expert Collaboration Diffusion Model2025-07-17From Variability To Accuracy: Conditional Bernoulli Diffusion Models with Consensus-Driven Correction for Thin Structure Segmentation2025-07-17Unleashing Vision Foundation Models for Coronary Artery Segmentation: Parallel ViT-CNN Encoding and Variational Fusion2025-07-17SCORE: Scene Context Matters in Open-Vocabulary Remote Sensing Instance Segmentation2025-07-17Unified Medical Image Segmentation with State Space Modeling Snake2025-07-17A Privacy-Preserving Semantic-Segmentation Method Using Domain-Adaptation Technique2025-07-17