Research map / Video / Video segmentation and tracking
Cubic LSTMs for Video Prediction
Hehe Fan, Linchao Zhu, Yi Yang
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 |
|---|---|---|---|
| separate spatial object processing from temporal motion processing | CubicLSTM + CDNA · Robotic Pushing | PSNR | Improved |
| replace ConvLSTM with CubicLSTM and use three spatial layers | CDNA · Robotic Pushing | PSNR | Improved |
| replace ConvLSTM with CNN, removing recurrent temporal state | CDNA · Robotic Pushing | PSNR | Hurt |
| use ConvLSTM instead of CNN | CDNA · Robotic Pushing | PSNR | No clear effect |
| use a three-spatial-layer rather than one-spatial-layer structure (ablation) | CubicRNN · Moving-MNIST MNIST-2 | MSE: 111.5 -> 73.4 | Improved |
| use a three-spatial-layer rather than one-spatial-layer structure (ablation) | CubicRNN · Moving-MNIST MNIST-3 | MSE: 158.4 -> 127.1 | Improved |
| use a three-spatial-layer rather than one-spatial-layer structure (ablation) | CubicRNN · Moving-MNIST MNIST-2 | BCE: 386.3 -> 210.3 | Improved |
| increase spatial layers from 2 to 3 (ablation) | CubicRNN (3 output layers) · Moving-MNIST MNIST-2 | MSE: 59.7 -> 47.3 | Improved |
| increase spatial layers from 2 to 3 (ablation) | CubicRNN (3 output layers) · Moving-MNIST MNIST-3 | MSE: 110.2 -> 88.2 | Improved |
| increase spatial layers from 2 to 3 (ablation) | CubicRNN (3 output layers) · Moving-MNIST MNIST-2 | BCE: 121.9 -> 91.7 | Improved |
| use 1 × 1 temporal convolution rather than 5 × 5 (ablation) | CubicLSTM · video prediction | prediction accuracy | Improved |
| replace ConvLSTM with CubicLSTM | KTH Action prediction · KTH Action | PSNR | Improved |
Closest papers by meaning
- Video Prediction Recalling Long-Term Motion Context via Memory Alignment Learning (CVPR 2021 · similarity 0.87)
- Learning to Generate Long-term Future via Hierarchical Prediction (ICML 2017 · similarity 0.87)
- Decomposing Motion and Content for Natural Video Sequence Prediction (ICLR 2017 · similarity 0.86)
- Lattice Long Short-Term Memory for Human Action Recognition (ICCV 2017 · similarity 0.86)
- Folded Recurrent Neural Networks for Future Video Prediction (ECCV 2018 · similarity 0.86)
- Visual Representation Learning with Stochastic Frame Prediction (ICML 2024 · 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.
