Research map / Video / Video generation
G3AN: Disentangling Appearance and Motion for Video Generation
Check 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 G³AN instead of MoCoGAN | video generation · MUG | FID: 87.11 -> 67.12 | Helped |
| Use G³AN instead of MoCoGAN | video generation · UvA-NEMO | FID: 197.32 -> 119.22 | Helped |
| Use G³AN instead of MoCoGAN | video generation · Weizmann | FID: 92.18 -> 86.01 | Helped |
| Use G³AN instead of MoCoGAN | video generation · UCF101 | FID: 104.14 -> 91.21 | Helped |
| Use G³AN instead of MoCoGAN | video generation · UCF101 | IS: 3.06 -> 3.62 | Helped |
| Add separate spatial, temporal, and factorized attention streams (ablation) | G³AN · MUG, UvA-NEMO, Weizmann, UCF101 | FID: 117.10 -> 67.12; 164.04 -> 119.22; 252.97 -> 86.01; 127.09 -> 91.21 | Helped |
| Add separate appearance and motion streams (ablation) | G³AN · MUG, UvA-NEMO, Weizmann, UCF101 | FID: 113.44 -> 67.12; 159.54 -> 119.22; 176.73 -> 86.01; 120.17 -> 91.21 | Helped |
| Add the spatial appearance stream G_S (ablation) | G³AN · MUG, UvA-NEMO, Weizmann, UCF101 | FID: 109.87 -> 67.12; 129.84 -> 119.22; 141.06 -> 86.01; 117.19 -> 91.21 | Helped |
| Add the spatial appearance stream G_S (ablation) | G³AN · UCF101 | IS: 3.05 -> 3.62 | Helped |
| Add factorized spatio-temporal self-attention (ablation) | G³AN · MUG, UvA-NEMO, Weizmann, UCF101 | FID: 85.11 -> 67.12; 128.14 -> 119.22; 97.54 -> 86.01; 98.37 -> 91.21 | Helped |
| Add factorized spatio-temporal self-attention (ablation) | G³AN · UCF101 | IS: 3.44 -> 3.62 | Helped |
| Add the temporal motion stream G_T (ablation) | G³AN · MUG, UvA-NEMO, Weizmann, UCF101 | FID: 82.07 -> 67.12; 121.87 -> 119.22; 94.64 -> 86.01; 96.47 -> 91.21 | Helped |
Closest papers by meaning
- Synergizing Motion and Appearance: Multi-Scale Compensatory Codebooks for Talking Head Video Generation (CVPR 2025 · similarity 0.87)
- VideoJAM: Joint Appearance-Motion Representations for Enhanced Motion Generation in Video Models (ICML 2025 · similarity 0.87)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.86)
- VideoSPatS: Video SPatiotemporal Splines for Disentangled Occlusion, Appearance and Motion Modeling and Editing (CVPR 2025 · similarity 0.85)
- MimicTalk: Mimicking a personalized and expressive 3D talking face in minutes (NeurIPS 2024 · similarity 0.85)
- OpenHumanVid: A Large-Scale High-Quality Dataset for Enhancing Human-Centric Video Generation (CVPR 2025 · 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.