Research map / Video / Video segmentation and tracking
HTML: Hybrid Temporal-scale Multimodal Learning Framework for Referring Video Object Segmentation
Mingfei Han, Yali Wang, Zhihui Li, Lina Yao, Xiaojun Chang, Yu Qiao
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add hybrid temporal scales (without CMP) to baseline, 8 frames (ablation) | ReferFormer ResNet-50 · Ref-YouTube-VOS | J&F: 56.2 -> 57.1 | Improved |
| add cross-scale multimodal perception (CMP) to HTML, 8 frames (ablation) | HTML · Ref-YouTube-VOS | J&F: 57.1 -> 57.8 | Improved |
| add hybrid temporal scales (without CMP), 5 frames (ablation) | ReferFormer ResNet-50 · Ref-YouTube-VOS | J&F: 55.6 -> 56.0 | Improved |
| add CMP to HTML, 5 frames (ablation) | HTML · Ref-YouTube-VOS | J&F: 56.0 -> 56.3 | Improved |
| use 3 temporal scales instead of 2 (ablation) | HTML · Ref-YouTube-VOS | J&F: 56.9 -> 57.8 | Improved |
| use 2 temporal scales instead of 1 (ablation) | HTML · Ref-YouTube-VOS | J&F: 56.2 -> 56.9 | Improved |
| increase input frames from 5 to 8 (ablation) | HTML · Ref-YouTube-VOS | J&F: 56.3 -> 57.8 | Improved |
| increase input frames from 8 to 12 (ablation) | HTML · Ref-YouTube-VOS | J&F: 57.8 -> 57.5 | No clear effect |
| use inter-scale perception l+1 -> l instead of l -> l+1 (ablation) | HTML · Ref-YouTube-VOS | J&F: 56.9 -> 57.8 | Improved |
| use 8 frames instead of 5 for baseline on short queries (ablation) | single-scale baseline · Ref-YouTube-VOS Short set | J&F: 55.6 -> 55.1 | Hurt |
| use Long instead of Short queries with 8 frames (ablation) | single-scale baseline · Ref-YouTube-VOS | J&F: 55.1 -> 55.9 | Improved |
| replace ReferFormer with HTML (ResNet-50) | ReferFormer · Ref-YouTube-VOS | J&F: 55.6 -> 57.8 | Improved |
Closest papers by meaning
- End-to-End Referring Video Object Segmentation With Multimodal Transformers (CVPR 2022 · similarity 0.92)
- Multi-Level Representation Learning With Semantic Alignment for Referring Video Object Segmentation (CVPR 2022 · similarity 0.90)
- DTOS: Dynamic Time Object Sensing with Large Multimodal Model (CVPR 2025 · similarity 0.89)
- LoSh: Long-Short Text Joint Prediction Network for Referring Video Object Segmentation (CVPR 2024 · similarity 0.89)
- Semantic and Sequential Alignment for Referring Video Object Segmentation (CVPR 2025 · similarity 0.88)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 2024 · similarity 0.88)
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.
