Research map / Video / Video generation
HECTOR: Hybrid Editable Compositional Object References for Video Generation
Guofeng Zhang, Angtian Wang, Jacob Zhiyuan Fang, Liming Jiang, Haotian Yang, Alan Yuille, Chongyang Ma
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 point-tracked trajectory scale instead of bounding-box scale (ablation) | HECTOR · DAVIS multi-object | centroid distance: 0.192 -> 0.167 | Improved |
| use point-tracked trajectory scale instead of bounding-box scale (ablation) | HECTOR · DAVIS multi-object | R-DINO: 0.4085 -> 0.4347 | Improved |
| add video references to training (hybrid image and video) (ablation) | HECTOR · DAVIS multi-object | centroid distance: 0.171 -> 0.167 | Improved |
| add video references to training (ablation) | HECTOR · DAVIS multi-object | CLIP-T: 0.3427 -> 0.3432 | No clear effect |
| use Gaussian-softened masks instead of binary masks (ablation) | HECTOR · DAVIS multi-object | R-DINO: 0.4211 -> 0.4347 | Improved |
| use Gaussian-softened masks instead of binary masks (ablation) | HECTOR · DAVIS multi-object | CLIP-T: 0.3315 -> 0.3432 | Improved |
| HECTOR trajectory-aligned references vs VACE (bbox) | VACE (bbox) · DAVIS single-object | mIoU: 0.2191 -> 0.3912 | Improved |
| HECTOR vs VACE (bbox) on text alignment | VACE (bbox) · DAVIS single-object | CLIP-T: 0.3363 -> 0.3306 | No clear effect |
| HECTOR vs VACE (mask) on R-DINO | VACE (mask) · DAVIS multi-object | R-DINO: 0.2540 -> 0.4347 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MotionBooth — MotionBooth: Motion-Aware Customized Text-to-Video Generation (NeurIPS 2024) · 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
- BlobGEN-Vid: Compositional Text-to-Video Generation with Blob Video Representations (CVPR 2025 · similarity 0.87)
- CAGE: Unsupervised Visual Composition and Animation for Controllable Video Generation (AAAI 2025 · similarity 0.87)
- Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (NeurIPS 2025 · similarity 0.87)
- UniReal: Universal Image Generation and Editing via Learning Real-world Dynamics (CVPR 2025 · similarity 0.87)
- AnimateAnything: Consistent and Controllable Animation for Video Generation (CVPR 2025 · similarity 0.86)
- Frame In-N-Out: Unbounded Controllable Image-to-Video Generation (NeurIPS 2025 · 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.
