Research map / Label-efficient and robust learning / Knowledge distillation
Distilled Siamese Networks for Visual Tracking
Jianbing Shen, Yuanpei Liu, Xingping Dong, Xiankai Lu, Fahad Shahbaz Khan, Steven C. H. Hoi
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 TS (teacher-student) loss to hard-label training (ablation) | DSTrpn student1 · OTB-100 | AUC: 0.429 -> 0.586 | Improved |
| add GT hard labels on top of TS loss (ablation) | DSTrpn student1 · OTB-100 | AUC: 0.586 -> 0.579 | Hurt |
| replace GT with AH loss in GT+TS (ablation) | DSTrpn student1 · OTB-100 | AUC: 0.579 -> 0.591 | Improved |
| add STR loss to TS (ablation) | DSTrpn student1 · OTB-100 | AUC: 0.586 -> 0.608 | Improved |
| use AH+TS+STR full transfer instead of TS+STR (ablation) | DSTrpn student1 · OTB-100 | AUC: 0.608 -> 0.624 | Improved |
| add STR loss to TS (ablation) | DSTfc student1 · OTB-100 | AUC: 0.535 -> 0.548 | Improved |
| add STR weighting strategy (ablation) | DSTrpn student1 (AH+TS+STR) · OTB-100 | AUC: 0.609 -> 0.624 | Improved |
| use proposed distillation loss instead of bounded loss (ablation) | DSTrpn student1 · OTB-100 | AUC: 0.612 -> 0.624 | Improved |
| add knowledge sharing from intelligent student (TSsKD vs TSKD) (ablation) | DSTrpn dim student · OTB-100 | AUC: 0.624 -> 0.646 | Improved |
| add knowledge sharing from intelligent student (TSsKD vs TSKD) (ablation) | DSTfc dim student · OTB-100 | AUC: 0.557 -> 0.573 | Improved |
| add distillation (TSKD) to intelligent student trained on hard labels (ablation) | SiamRPN student2 · OTB-100 | AUC: 0.630 -> 0.641 | Improved |
| train students with 25% gray images instead of 0% (ablation) | DSTrpn · OTB-100 | AUC: 0.621 -> 0.646 | Improved |
Closest papers by meaning
- SiamRPN++: Evolution of Siamese Visual Tracking With Very Deep Networks (CVPR 2019 · similarity 0.89)
- SiamMOT: Siamese Multi-Object Tracking (CVPR 2021 · similarity 0.89)
- Deformable Siamese Attention Networks for Visual Object Tracking (CVPR 2020 · similarity 0.88)
- Deeper and Wider Siamese Networks for Real-Time Visual Tracking (CVPR 2019 · similarity 0.88)
- Adaptive Siamese Tracking with a Compact Latent Network (TPAMI 2022 · similarity 0.88)
- High Performance Visual Tracking With Siamese Region Proposal Network (CVPR 2018 · 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.
