Research map / Video / Video segmentation and tracking
A polar prediction model for learning to represent visual transformations
Pierre-Étienne H Fiquet, Eero P Simoncelli
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 multiscale polar predictor instead of copying last frame | Next-frame prediction · DAVIS test | PSNR (dB): 20.02 -> 24.11 | Improved |
| use mPP instead of causal motion compensation (cMC) | Next-frame prediction · DAVIS test | PSNR (dB): 22.37 -> 24.11 | Improved |
| use mPP instead of steerable pyramid phase extrapolation (SPyr) | Next-frame prediction · DAVIS test | PSNR (dB): 23.82 -> 24.11 | Improved |
| use mPP instead of CNN with 20 nonlinear stages | Next-frame prediction · DAVIS test | PSNR (dB): 23.58 -> 24.11 | Improved |
| use mPP instead of Unet | Next-frame prediction · DAVIS test | PSNR (dB): 23.94 -> 24.11 | Improved |
| use quadratic predictor (mQP) instead of polar predictor (mPP) | mPP · DAVIS test | PSNR (dB): 24.11 -> 24.04 | No clear effect |
| use mQP instead of mPP | mPP · VanHateren test | PSNR (dB): 29.11 -> 29.07 | No clear effect |
| compare mPP with Unet | Unet · VanHateren test | PSNR (dB): 29.26 -> 29.11 | Hurt |
| compare mPP with CNN | CNN · VanHateren test | PSNR (dB): 29.09 -> 29.11 | No clear effect |
| use mQP instead of mPP for computation | mPP · VanHateren | training time (min:sec): 10:24 -> 46:14 | Hurt |
| use mPP instead of CNN in parameter count | CNN · Model size | # parameters: 666,496 -> 7,744 | Improved |
| untie analysis and synthesis transforms (ablation) | Polar predictor (single scale) · VanHateren test | PSNR (dB): 28.83 -> 28.87 | No clear effect |
Closest papers by meaning
- Learning predictable and robust neural representations by straightening image sequences (NeurIPS 2024 · similarity 0.88)
- NEvo: Neural-Guided Evolutionary Video Synthesis for Dynamic Visual Selectivity (NeurIPS 2026 · similarity 0.86)
- Visual Representation Learning with Stochastic Frame Prediction (ICML 2024 · similarity 0.86)
- Contrastive-Equivariant Self-Supervised Learning Improves Alignment with Primate Visual Area IT (NeurIPS 2024 · similarity 0.86)
- Time-Evolving Dynamical System for Learning Latent Representations of Mouse Visual Neural Activity (NeurIPS 2025 · similarity 0.85)
- Flexible Context-Driven Sensory Processing in Dynamical Vision Models (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.
