AI Paper Atlas

Research map / Video / Video segmentation and tracking

Few-Shot Referring Video Single- and Multi-Object Segmentation Via Cross-Modal Affinity with Instance Sequence Matching

Heng Liu, Guanghui Li, Mingqi Gao, Xiantong Zhen, Feng Zheng, Yang Wang

IJCV 2025 · Video › Video segmentation and tracking · approach: Mask propagation for unsupervised video segmentation · 11 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add self-affinity block to baseline (ablation)CMA · Mini-Ref-YouTube-VOS Fold-1J&F: 57.9 -> 59.2Improved
add cross-affinity block on top of self-affinity (ablation)CMA · Mini-Ref-YouTube-VOS Fold-1J&F: 59.2 -> 60.2Improved
add instance sequence matching (ISM) (ablation)Few-shot RVOS model · Mini-MeViS Fold-3J&F: 38.2 -> 38.7Improved
add text expression for unseen categories (ablation)Few-shot RVOS model · Bottle (unseen)J: 77.3 -> 82.4Improved
add text expression for unseen categories (ablation)Few-shot RVOS model · Lego (unseen)J: 69.5 -> 84.2Improved
add text expression for unseen categories (ablation)Few-shot RVOS model · Coffee (unseen)F: 64.3 -> 71.0Improved
add text expression for unseen categories (ablation)Few-shot RVOS model · Oven (unseen)F: 31.4 -> 35.6Improved
vary matching threshold sigma (ablation)ISM · Mini-Ref-YouTube-VOS Fold-1J&FNo clear effect
fine-tune on 18 Mini-MeViS videosFew-shot RVOS model (Fold3) · Mini-MeViSJ&F: 38.7 -> 41.6Improved
fine-tune RVOS baseline ReferFormer on Mini-MeViS videosReferFormer · Mini-MeViSJ&F: 38.2 -> 40.6Improved
fine-tune on Mini-Ref-SAIL-VOSMUTR · Mini-Ref-SAIL-VOSJ&F: 68.4 -> 72.9Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy