Research map / Video / Video segmentation and tracking
Stochastic Latent Residual Video Prediction
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 latent residual dynamics instead of SVG's image-autoregressive model | Stochastic video prediction · Moving MNIST stochastic | PSNR: 14.50 ± 0.04 -> 16.93 ± 0.07 | Helped |
| Use latent residual dynamics instead of SVG's image-autoregressive model | Stochastic video prediction · Moving MNIST deterministic | PSNR: 12.85 ± 0.03 -> 18.25 ± 0.06 | Helped |
| Use latent residual dynamics instead of the strongest reported baseline | Stochastic video prediction · KTH | LPIPS: 0.2049 ± 0.0053 -> 0.0736 ± 0.0029 | Helped |
| Use latent residual dynamics instead of StructVRNN | Stochastic video prediction · Human3.6M | PSNR: 24.46 ± 0.17 -> 25.30 ± 0.19 | Helped |
| Use the proposed model rather than SV2P | Stochastic video prediction · BAIR | LPIPS: 0.0912 ± 0.0053 -> 0.0574 ± 0.0032 | Helped |
| Replace residual dynamics with an MLP (ablation) | Ours · Moving MNIST stochastic | PSNR: 16.93 ± 0.07 -> 16.55 ± 0.06 | Hurt |
| Replace residual dynamics with a GRU (ablation) | Ours · Moving MNIST deterministic | PSNR: 18.25 ± 0.06 -> 13.17 ± 0.03 | Hurt |
| Remove stochastic latent variable z (ablation) | Ours · Moving MNIST deterministic | PSNR: 18.25 ± 0.06 -> 14.99 ± 0.03 | Hurt |
| Halve the Euler step size at test time (ablation) | Ours · KTH | PSNR: 29.69 ± 0.32 -> 29.43 ± 0.33 | No clear effect |
| Replace residual dynamics with an MLP (ablation) | Ours · Human3.6M | FVD: 416 ± 5 -> 582 ± 4 | Hurt |
| Replace residual dynamics with a GRU (ablation) | Ours · Human3.6M | FVD: 416 ± 5 -> 1050 ± 20 | Hurt |
| Halve the Euler step size at test time (ablation) | Ours · BAIR | PSNR: 19.59 ± 0.27 -> 19.45 ± 0.26 | No clear effect |
Closest papers by meaning
- STDiff: Spatio-Temporal Diffusion for Continuous Stochastic Video Prediction (AAAI 2024 · similarity 0.89)
- TRIP: Temporal Residual Learning with Image Noise Prior for Image-to-Video Diffusion Models (CVPR 2024 · similarity 0.87)
- Video Prediction by Modeling Videos as Continuous Multi-Dimensional Processes (CVPR 2024 · similarity 0.86)
- DFDNet: Disentangling and Filtering Dynamics for Enhanced Video Prediction (AAAI 2025 · similarity 0.85)
- Efficient Video Diffusion Models via Content-Frame Motion-Latent Decomposition (ICLR 2024 · similarity 0.85)
- Generative Pre-trained Autoregressive Diffusion Transformer (NeurIPS 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.