Research map / Video / Video segmentation and tracking
Temporal Coherency based Criteria for Predicting Video Frames using Deep Multi-stage Generative Adversarial 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 |
|---|---|---|---|
| add PCDL and higher-order PCDL losses to Adv + NCCL + L1 (Combined) (ablation) | Multi-stage GAN · UCF-101 1st frame | PSNR/SSIM: 35.4/0.94 -> 37.3/0.95 | Helped |
| add PCDL and higher-order PCDL losses to Adv + NCCL + L1 (ablation) | Multi-stage GAN · UCF-101 4th frame | PSNR/SSIM: 28.7/0.75 -> 30.2/0.76 | Helped |
| add PCDL and higher-order PCDL losses to Adv + NCCL + L1 (ablation) | Multi-stage GAN · KITTI 1st frame | PSNR/SSIM: 37.1/0.91 -> 39.7/0.93 | Helped |
| add SNCCL to Combined loss (ablation) | Multi-stage GAN · UCF-101 1st frame | PSNR/SSIM: 37.3/0.95 -> 38.2/0.95 | Helped |
| add SNCCL to Combined loss (ablation) | Multi-stage GAN · UCF-101 4th frame | PSNR/SSIM: 30.2/0.76 -> 30.9/0.77 | Helped |
| evaluate on full frames instead of motion patches | Combined + SNCCL · UCF-101 1st frame | PSNR/SSIM: 38.2/0.95 -> 37.3/0.94 | Hurt |
| use multi-stage GAN with Combined + SNCCL instead of Deep Voxel Flow | Combined + SNCCL vs Deep Voxel Flow · UCF-101 1st frame | PSNR/SSIM: 35.8/0.96 -> 38.2/0.95 | Mixed |
| use temporal-coherency losses (NCCL, PCDL) instead of L1/GDL-based adversarial training | Multi-stage GAN · UCF-101 4th frame | PSNR/SSIM | Helped |
Closest papers by meaning
- Predicting Future Frames Using Retrospective Cycle GAN (CVPR 2019 · similarity 0.91)
- Learning to Generate Time-Lapse Videos Using Multi-Stage Dynamic Generative Adversarial Networks (CVPR 2018 · similarity 0.89)
- Generating Videos with Scene Dynamics (NeurIPS 2016 · similarity 0.88)
- Visual Dynamics: Probabilistic Future Frame Synthesis via Cross Convolutional Networks (NeurIPS 2016 · similarity 0.87)
- Towards Smooth Video Composition (ICLR 2023 · similarity 0.86)
- Video Prediction by Modeling Videos as Continuous Multi-Dimensional Processes (CVPR 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.