Research map / Vision-language and multimodal / CLIP and zero-shot
Visual Prompt Multi-Modal Tracking
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 full fine-tuning with frozen-backbone visual prompt tuning (MCP blocks, 0.84M trainable params) (ablation) | ViPT vs FFT (auxiliary-modal branch full fine-tuning) · DepthTrack / LasHeR / VisEvent test | F-score / SR / SR: 59.4 vs 55.6 F-score; 52.5 vs 51.7 SR; 59.2 vs 58.7 SR | Helped |
| Add auxiliary-modal input via prompt learning vs RGB-only foundation tracker (ablation) | ViPT (multi-modal) vs OSTrack RGB foundation · DepthTrack / LasHeR / VisEvent | F-score / SR / SR: +6.5 F-score; +11.3 SR; +5.8 SR | Helped |
| Use VPT-style deep prompts instead of MCP blocks (ablation) | Prompt-deep vs ViPT · DepthTrack / LasHeR | F-score / SR: 54.9 vs 59.4 F-score; 50.7 vs 52.5 SR | Hurt |
| Insert MCP blocks only in shallow layers (ViPT-shaw) instead of deep placement (ablation) | ViPT-shaw vs ViPT · DepthTrack / LasHeR / VisEvent | F-score / SR: 56.4 vs 59.4 F-score; 47.9 vs 52.5 SR; 57.8 vs 59.2 SR | Hurt |
| Expand training data with generated RGB-D/RGB-T pairs (ablation) | ViPT · DepthTrack / LasHeR | F-score / SR: 59.3 vs 59.4 F-score (no gain); 49.7 vs 52.5 SR | Hurt |
| Unfreeze all parameters (full tuning) instead of prompt-only tuning (ablation) | ViPT · DepthTrack / LasHeR | F-score / SR: 55.4 vs 59.4 F-score; 51.3 vs 52.5 SR | Hurt |
| Combine expanded data and unfrozen parameters (ablation) | ViPT · LasHeR | SR: 53.9 vs 52.5 SR (gain) but 58.3 vs 59.4 F-score (loss) on DepthTrack | Mixed |
| Insert MCP blocks at more layers (denser placement) (ablation) | ViPT · DepthTrack / LasHeR / VisEvent | performance on all benchmarks | Helped |
| Add spatial fovea and MCP prompter with RGB-branch prompts vs multi-modal prompts in ViPT | ViPT on DepthTrack · DepthTrack test | F-score: 59.4 vs 52.9 for foundation | Helped |
| Prompt-tuning vs ProTrack-style training-free prompt fusion | ViPT vs ProTrack · RGBT234 MSR / VOT-RGBD2022 EAO | MSR / EAO: 61.7 vs 59.9 MSR; 0.721 vs 0.651 EAO | Helped |
Closest papers by meaning
- Collaborating Vision, Depth, and Thermal Signals for Multi-Modal Tracking: Dataset and Algorithm (NeurIPS 2025 · similarity 0.89)
- Bi-directional Adapter for Multimodal Tracking (AAAI 2024 · similarity 0.89)
- SDSTrack: Self-Distillation Symmetric Adapter Learning for Multi-Modal Visual Object Tracking (CVPR 2024 · similarity 0.88)
- Single-Model and Any-Modality for Video Object Tracking (CVPR 2024 · similarity 0.86)
- Explicit Visual Prompts for Visual Object Tracking (AAAI 2024 · similarity 0.86)
- Facing the Elephant in the Room: Visual Prompt Tuning or Full finetuning? (ICLR 2024 · 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.