Research map / Video / Video segmentation and tracking
Decomposing Motion and Content for Natural Video Sequence 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 |
|---|---|---|---|
| Separate motion and content into dedicated encoder pathways | MCnet · KTH | SSIM and PSNR | Helped |
| Separate motion and content into dedicated encoder pathways | MCnet · Weizmann action | SSIM and PSNR | Helped |
| Separate motion and content into dedicated encoder pathways | MCnet · UCF-101 | image quality | Helped |
| Train on Sports-1M rather than UCF-101 | MCnet + RES · UCF-101 | image quality | No clear effect |
| Use last-frame copying instead of learned prediction | MCnet on UCF-101 · UCF-101 | overall image quality | Hurt |
| Evaluate only pixels where optical-flow motion is observed | MCnet versus baselines · UCF-101 motion regions | SSIM | Helped |
| Evaluate only pixels where optical-flow motion is observed | MCnet versus copy-last-frame baseline · UCF-101 motion regions, first few prediction steps | PSNR | Helped |
| Evaluate only pixels where optical-flow motion is observed | MCnet versus copy-last-frame baseline · UCF-101 motion regions, later prediction steps | PSNR | Hurt |
| Test on videos with the greatest overall motion | MCnet versus baselines · UCF-101, highest-motion deciles | image quality | Helped |
| Add multi-scale motion-content residual connections (ablation) | MCnet · KTH and Weizmann action | SSIM and PSNR | Helped |
| Add multi-scale motion-content residual connections (ablation) | ConvLSTM baseline · KTH and Weizmann action | SSIM and PSNR | Mixed |
Closest papers by meaning
- Learning to Generate Long-term Future via Hierarchical Prediction (ICML 2017 · similarity 0.89)
- Flexible Spatio-Temporal Networks for Video Prediction (CVPR 2017 · similarity 0.89)
- Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning (ICLR 2017 · similarity 0.89)
- Structure Preserving Video Prediction (CVPR 2018 · similarity 0.87)
- Convolutional Sequence to Sequence Model for Human Dynamics (CVPR 2018 · similarity 0.87)
- Visual Dynamics: Probabilistic Future Frame Synthesis via Cross Convolutional Networks (NeurIPS 2016 · similarity 0.87)
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.