Research map / Video / Video segmentation and tracking
Joint Inductive and Transductive Learning for Video Object Segmentation
Yunyao Mao, Ning Wang, Wengang Zhou, Houqiang Li
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 the induction branch to the transduction branch (ablation) | JOINT · YouTube-VOS 2018 val | Overall: 81.2 -> 83.1 | Improved |
| add the induction branch to the transduction branch (ablation) | JOINT · YouTube-VOS 2018 val unseen categories | J unseen: 75.1 -> 78.7 | Improved |
| add the induction branch to the transduction branch (ablation) | JOINT · YouTube-VOS 2018 val unseen categories | F unseen: 83.2 -> 86.5 | Improved |
| add the transduction branch to the induction branch (LWL) (ablation) | JOINT · YouTube-VOS 2018 val | Overall: 81.5 -> 83.1 | Improved |
| add the transduction branch to the induction branch (LWL) (ablation) | JOINT · YouTube-VOS 2018 val | seconds per frame: 0.15 -> 0.25 | Hurt |
| use a two-head label encoder instead of single-head (ablation) | JOINT · YouTube-VOS 2018 val | Overall: 81.8 -> 82.4 | Improved |
| add the cosine similarity disentangling loss to the two-head encoder (ablation) | JOINT · YouTube-VOS 2018 val | Overall: 82.4 -> 83.1 | Improved |
| add the cosine disentangling loss (ablation) | JOINT · YouTube-VOS 2018 val unseen | J unseen: 77.6 -> 78.7 | Improved |
| add the cosine disentangling loss (ablation) | JOINT · YouTube-VOS 2018 val seen | J seen: 81.1 -> 81.5 | Improved |
| use the joint model rather than the induction-only LWL on DAVIS 2017 (YouTube-VOS training) | JOINT · DAVIS 2017 val | J&F: 81.6 -> 83.5 | Improved |
| use the joint model rather than the induction-only LWL on DAVIS 2017 (DAVIS only training) | JOINT · DAVIS 2017 val (DAVIS-only) | J&F: 74.3 -> 78.6 | Improved |
| compare seen-category score against multi-scale CFBI | JOINT · YouTube-VOS 2018 val seen | J seen: 82.2 -> 81.5 | Hurt |
Closest papers by meaning
- A Transductive Approach for Video Object Segmentation (CVPR 2020 · similarity 0.92)
- Learning to Learn Better for Video Object Segmentation (AAAI 2023 · similarity 0.91)
- Temporal Transductive Inference for Few-Shot Video Object Segmentation (IJCV 2025 · similarity 0.88)
- Fast Video Object Segmentation by Reference-Guided Mask Propagation (CVPR 2018 · similarity 0.87)
- From ViT Features to Training-free Video Object Segmentation via Streaming-data Mixture Models (NeurIPS 2023 · similarity 0.87)
- Video Object Segmentation without Temporal Information (TPAMI 2018 · similarity 0.87)
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.
