Research map / Detection and segmentation / Open-set and few-shot detection
Isomer: Isomerous Transformer for Zero-shot Video Object Segmentation
Yichen Yuan, Yifan Wang, Lijun Wang, Xiaoqi Zhao, Huchuan Lu, Yu Wang, Weibo Su, Lei Zhang
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 |
|---|---|---|---|
| replace vanilla Transformer with CST in the first two fusion stages (ablation) | Isomer (MiT-b0) · DAVIS-16 val | FPS / J Mean: 3 FPS -> 36 FPS, J Mean 84.2 -> 84.2 | Improved |
| replace vanilla Transformer with SGST in the last two fusion stages (ablation) | Isomer (MiT-b0) · DAVIS-16 val | J Mean: 84.2 -> 84.6, 36 -> 39 FPS | Improved |
| use CST in high-level stages instead of SGST (ablation) | Isomer (MiT-b0) · DAVIS-16 val | J Mean: 84.6 -> 82.9 | Hurt |
| level-isomerous scheme (CST low, SGST high) rather than SGST everywhere (ablation) | Isomer (MiT-b0) · DAVIS-16 val | FPS: 29 -> 39 FPS, J Mean 84.5 -> 84.6 | Improved |
| foreground-background separate modeling in SGST (ablation) | SGST · DAVIS-16 val | J Mean: 84.3 -> 84.6 | Improved |
| model only foreground in SGST (ablation) | SGST · DAVIS-16 val | J Mean: 84.6 -> 81.8 | Hurt |
| soft merging ratio K/N of 1/9 versus no merging (ablation) | SGST · DAVIS-16 val | J Mean: 83.6 -> 84.6 | Improved |
| merge too aggressively (K/N 1/36) (ablation) | SGST · DAVIS-16 val | J Mean: 84.6 -> 83.7 | Hurt |
| use CST/SGST instead of generic lightweight Transformers as fusion | Isomer (MiT-b0) · DAVIS-16 val | J Mean: 84.1 (EdgeViT) -> 84.6 | Improved |
| Isomer fusion versus HFAN across backbones | Isomer · DAVIS-16 val | J&F: ResNet101 87.0 -> 87.5; Swin-Tiny 86.7 -> 89.2 | Improved |
Closest papers by meaning
- Adaptive Multi-Source Predictor for Zero-Shot Video Object Segmentation (IJCV 2024 · similarity 0.88)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 2024 · similarity 0.87)
- Reciprocal Transformations for Unsupervised Video Object Segmentation (CVPR 2021 · similarity 0.87)
- OneVOS: Unifying Video Object Segmentation with All-in-One Transformer Framework (ECCV 2024 · similarity 0.86)
- DiffCut: Catalyzing Zero-Shot Semantic Segmentation with Diffusion Features and Recursive Normalized Cut (NeurIPS 2024 · similarity 0.86)
- VidSeg: Training-free Video Semantic Segmentation based on Diffusion Models (CVPR 2025 · similarity 0.86)
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.
