Research map / Video / Video segmentation and tracking
High Fidelity Video Prediction with Large Stochastic Recurrent Neural Networks
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 |
|---|---|---|---|
| Increase capacity from M=1, K=1 to M=3, K=5 (ablation) | SVG' · Towel Pick | FVD: 189.91 -> 93.71 | Helped |
| Increase capacity from M=1, K=1 to M=3, K=5 (ablation) | SVG' · Human 3.6M | FVD: 682.08 -> 429.88 | Helped |
| Increase capacity from M=1, K=1 to M=3, K=5 (ablation) | SVG' · KITTI | FVD: 2264.91 -> 1217.25 | Helped |
| Add recurrent connections (ablation) | CNN video prediction · Towel Pick | FVD: 199.81 -> 100.04 | Helped |
| Add recurrent connections (ablation) | CNN video prediction · Human 3.6M | FVD: 1321.23 -> 458.77 | Helped |
| Add recurrent connections (ablation) | CNN video prediction · KITTI | FVD: 2414.64 -> 1159.25 | Helped |
| Add stochastic latent variables (ablation) | LSTM video prediction · Towel Pick | FVD: 100.04 -> 93.71 | Helped |
| Add stochastic latent variables (ablation) | LSTM video prediction · Human 3.6M | FVD: 458.77 -> 429.88 | Helped |
| Add stochastic latent variables (ablation) | LSTM video prediction · KITTI | FVD: 1159.25 -> 1217.25 | Hurt |
| Use 5 context frames instead of 2 (ablation) | SVG' · Human 3.6M | Cosine similarity: 0.916 -> 0.925 | Helped |
| Use 10 rather than 5 context frames (ablation) | SVG' · Human 3.6M | Cosine similarity: 0.925 -> 0.925 | No clear effect |
| Use skip connections (ablation) | SVG' and LSTM video prediction · Towel Pick | FVD | Helped |
Closest papers by meaning
- Flexible Spatio-Temporal Networks for Video Prediction (CVPR 2017 · similarity 0.88)
- Visual Representation Learning with Stochastic Frame Prediction (ICML 2024 · similarity 0.86)
- Frame Context Packing and Drift Prevention in Next-Frame-Prediction Video Diffusion Models (NeurIPS 2025 · similarity 0.85)
- Bias for Action: Video Implicit Neural Representations with Bias Modulation (CVPR 2025 · similarity 0.85)
- STDiff: Spatio-Temporal Diffusion for Continuous Stochastic Video Prediction (AAAI 2024 · similarity 0.85)
- Motion Graph Unleashed: A Novel Approach to Video Prediction (NeurIPS 2024 · 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.