Research map / Vision-language and multimodal / Cross-modal learning
TrajPrompt: Aligning Color Trajectory with Vision-Language Representations
Li-Wu Tsao, Hao-Tang Tsui, Yu-Rou Tuan, Pei-Chi Chen, Kuan-Lin Wang, Jhih-Ciang Wu, Hong-Han Shuai, Wen-Huang Cheng
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 image-text contrastive loss (ablation) | TrajPrompt · SDD | ADE: 10.15 -> 3.78 | Improved |
| add alignment loss (ablation) | TrajPrompt · SDD | ADE: 5.12 -> 3.78 | Improved |
| add surrounding text embeddings to the T2T sequence (ablation) | TrajPrompt · SDD | ADE: 5.21 -> 3.78 | Improved |
| add conditional image prompt (ablation) | TrajPrompt · SDD | ADE: 4.30 -> 3.78 | Improved |
| draw trajectory with Gaussian instead of square points (ablation) | TrajPrompt colored line · SDD | ADE: 7.33 -> 3.78 | Improved |
| draw trajectory with Gaussian instead of circle points (ablation) | TrajPrompt colored line · SDD | ADE: 4.28 -> 3.78 | Improved |
| use image-text contrastive loss with semantic text instead of PreTraM map-trajectory contrastive loss (ablation) | TrajPrompt · SDD | ADE: 7.90 -> 4.69 | Improved |
| retrieve nearest pixel embedding from dictionary at each decoding step (ablation) | TrajPrompt decoding · SDD | ADE: 17.03 -> 7.73 | Improved |
| use frozen CLIP with soft prompts instead of full trajectory models | TrajPrompt · SDD | ADE: 6.52 -> 3.78 | Improved |
| use TrajPrompt instead of TDOR | DroneCrowd · DroneCrowd | ADE: 6.92 -> 3.72 | Improved |
Closest papers by meaning
- Trajectory-aware Cross-view Geo-Localization with Sequential Observations (ECCV 2026 · similarity 0.87)
- VisionTrap: Vision-Augmented Trajectory Prediction Guided by Textual Descriptions (ECCV 2024 · similarity 0.87)
- TrajRAG: Retrieving Geometric-Semantic Experience for Zero-Shot Object Navigation (CVPR 2026 · similarity 0.86)
- Unveiling the Value of Motion for Cinematic Camera Trajectories (NeurIPS 2026 · similarity 0.86)
- MolmoMotion: Forecasting Point Trajectories in 3D with Language Instruction (NeurIPS 2026 · similarity 0.85)
- TrajLoc: Trajectory-Attention Localization for Multi-Object Motion Control (NeurIPS 2026 · similarity 0.85)
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.
