Research map / Video / Video generation
Compositional Video Synthesis with Action Graphs
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 |
|---|---|---|---|
| use clocked action edges to track timed action progress | AG2Vid · CATER action-timing evaluation | human timing accuracy: 89.45 | Helped |
| replace the LGF GCN with an edge-sequence RNN (ablation) | AG2Vid layout generation on CATER · CATER | mIOU: 93.09 vs. 75.71 | Hurt |
| replace the LGF GCN with an edge-sequence RNN (ablation) | AG2Vid layout generation on SmthV2 · SmthV2 | mIOU: 51.32 vs. 41.28 | Hurt |
| add perceptual and feature matching losses to flow loss (ablation) | AG2Vid frame generation · SmthV2 | Inception Score: 1.59 ± 0.02 -> 2.21 ± 0.07 | Helped |
| add the pixel action discriminator loss (ablation) | AG2Vid frame generation · SmthV2 | Inception Score: 2.21 ± 0.07 -> 3.02 ± 0.11 | Helped |
| add perceptual and feature matching losses to flow loss (ablation) | AG2Vid frame generation · CATER | LPIPS: 0.14 ± 0.01 -> 0.08 ± 0.03 | Helped |
| add the pixel action discriminator loss (ablation) | AG2Vid frame generation · CATER | LPIPS: 0.08 ± 0.03 -> 0.07 ± 0.02 | Helped |
| use ground-truth layouts at inference (ablation) | AG2Vid on SmthV2 · SmthV2 | Inception Score: 3.02 ± 0.11 -> 3.52 ± 0.14 | Helped |
| use ground-truth layouts at inference (ablation) | AG2Vid on CATER · CATER | LPIPS: 0.07 ± 0.02 -> 0.06 ± 0.02 | Helped |
| use ground-truth layouts at inference (ablation) | AG2Vid on SmthV2 · SmthV2 | LPIPS: 0.25 ± 0.08 -> 0.22 ± 0.09 | Helped |
| compose learned actions into unseen action combinations | AG2Vid · CATER and SmthV2 zero-shot compositions | average class recall: 96.65 and 87.5 | Helped |
Closest papers by meaning
- T2V-CompBench: A Comprehensive Benchmark for Compositional Text-to-video Generation (CVPR 2025 · similarity 0.87)
- CAGE: Unsupervised Visual Composition and Animation for Controllable Video Generation (AAAI 2025 · similarity 0.86)
- VideoTetris: Towards Compositional Text-to-Video Generation (NeurIPS 2024 · similarity 0.86)
- Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis (CVPR 2024 · similarity 0.86)
- DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control (NeurIPS 2026 · similarity 0.86)
- Make Pixels Dance: High-Dynamic Video Generation (CVPR 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.