Research map / Video / Video segmentation and tracking
Clockwork Variational Autoencoders
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 |
|---|---|---|---|
| add temporal abstraction (clock speed factor 6 per level) to a 3-level latent hierarchy (ablation) | CW-VAE · All 4 datasets x 4 metrics | mean rank: 2.62 -> 2.25 | Helped |
| add temporal abstraction to hierarchy (ablation) | CW-VAE · MineRL Navigate | PSNR: 20.39 -> 23.23 | Helped |
| add temporal abstraction to hierarchy (ablation) | CW-VAE · KTH Action | PSNR: 25.21 -> 26.52 | Helped |
| add temporal abstraction to hierarchy (ablation) | CW-VAE · GQN Mazes | PSNR: 13.91 -> 13.71 | No clear effect |
| add temporal abstraction to hierarchy (ablation) | CW-VAE · Moving MNIST | PSNR: 12.93 -> 13.11 | Helped |
| add hierarchical latents without temporal abstraction (NoTmpAbs vs single-level RSSM) (ablation) | RSSM · All 4 datasets x 4 metrics | mean rank: 2.75 -> 2.62 | Helped |
| increase temporal abstraction factor from 2 to 8 (ablation) | CW-VAE · Moving MNIST | prediction quality over horizon | Helped |
| use deeper hierarchy (more levels) (ablation) | CW-VAE · Moving MNIST | Level 1 KL: 440.50 -> 397.10 | Helped |
| CW-VAE vs SVG feeding generated images back | CW-VAE · Moving MNIST | FVD: 1392 -> 593 | Helped |
| CW-VAE vs VTA two-level temporal abstraction | CW-VAE · MineRL Navigate | PSNR: 18.50 -> 23.23 | Helped |
Closest papers by meaning
- CV-VAE: A Compatible Video VAE for Latent Generative Video Models (NeurIPS 2024 · similarity 0.88)
- WF-VAE: Enhancing Video VAE by Wavelet-Driven Energy Flow for Latent Video Diffusion Model (CVPR 2025 · similarity 0.87)
- Improved Video VAE for Latent Video Diffusion Model (CVPR 2025 · similarity 0.87)
- Deep Hierarchical Video Compression (AAAI 2024 · similarity 0.85)
- VidTwin: Video VAE with Decoupled Structure and Dynamics (CVPR 2025 · similarity 0.85)
- WorldWeaver: Generating Long-Horizon Video Worlds via Rich Perception (NeurIPS 2025 · similarity 0.84)
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.