Research map / Video / Video generation
FlexTraj: Image-to-Video Generation with Flexible Point Trajectory Control
Zhiyuan Zhang, Can Wang, Dongdong Chen, Jing Liao
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 |
|---|---|---|---|
| use sequence-concatenation conditioning instead of ControlNet-style injection (ablation) | FlexTraj · DAVIS aligned | TrajErr: 0.131 -> 0.024 | Improved |
| use sequence-concatenation conditioning instead of ControlNet-style injection (ablation) | FlexTraj · DAVIS unaligned | TrajSIM: 0.556 -> 0.908 | Improved |
| use density/alignment annealing curriculum instead of random task mixing (ablation) | FlexTraj · DAVIS aligned | TrajErr: 0.126 -> 0.024 | Improved |
| use density/alignment annealing curriculum instead of random task mixing (ablation) | FlexTraj · DAVIS aligned | FVD: 1034.6 -> 693.3 | Improved |
| add trajectory ID to point representation (ablation) | FlexTraj · DAVIS aligned | TrajErr: 0.029 -> 0.024 | Improved |
| add trajectory ID to point representation (ablation) | FlexTraj · DAVIS aligned | FVD: 668.2 -> 693.3 | No clear effect |
| add segmentation ID to point representation (ablation) | FlexTraj · DAVIS aligned | TrajErr: 0.040 -> 0.024 | Improved |
| add segmentation ID to point representation (ablation) | FlexTraj · DAVIS aligned | FVD: 707.9 -> 693.3 | No clear effect |
| use trained point trajectory model with sparse input instead of densifying with templates (Sparse2Dense) (ablation) | FlexTraj · DAVIS aligned | TrajErr: 0.126 -> 0.024 | Improved |
| use FlexTraj over best baseline for dense control | FlexTraj · DAVIS dense | FVD: 705.3 -> 532.4 | Improved |
| use FlexTraj over DAS for dense control | FlexTraj · FlexBench dense | FVD: 1338.8 -> 1397.8 | Hurt |
| use FlexTraj with Wan2.2 5B vs Wan-Move 14B | FlexTraj · MoveBench | PSNR/EPE: 17.8 -> 17.4 PSNR; 2.6 -> 2.7 EPE | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- LeviTor — LeviTor: 3D Trajectory Oriented Image-to-Video Synthesis (CVPR 2025) · better on 2 of 2 shared results
- ToRA — Tora: Trajectory-oriented Diffusion Transformer for Video Generation (CVPR 2025) · better on 2 of 2 shared results
- Wan-Move — Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (NeurIPS 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (NeurIPS 2025 · similarity 0.90)
- MotionPro: A Precise Motion Controller for Image-to-Video Generation (CVPR 2025 · similarity 0.89)
- PoseTraj: Pose-Aware Trajectory Control in Video Diffusion (CVPR 2025 · similarity 0.88)
- MotionClone: Training-Free Motion Cloning for Controllable Video Generation (ICLR 2025 · similarity 0.88)
- SG-I2V: Self-Guided Trajectory Control in Image-to-Video Generation (ICLR 2025 · similarity 0.88)
- Motion Prompting: Controlling Video Generation with Motion Trajectories (CVPR 2025 · 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.
