Research map / Video / Video segmentation and tracking
Learning Joint Spatial-Temporal Transformations for Video Inpainting
Yanhong Zeng, Jianlong Fu, Hongyang Chao
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 multi-scale patches across heads instead of a single 108x60 patch size (ablation) | STTN · DAVIS | PSNR: 30.16 -> 30.67 | Improved |
| use multi-scale patches instead of a single 9x5 patch size (ablation) | STTN · DAVIS | VFID: 0.163 -> 0.149 | Improved |
| use multi-scale instead of single 36x20 patches (ablation) | STTN · DAVIS | E_warp: 0.2051 -> 0.1779 | Improved |
| stack 8 spatial-temporal transformer layers instead of 2 (ablation) | STTN · DAVIS | PSNR: 30.17 -> 30.67 | Improved |
| stack 8 layers instead of 4 (ablation) | STTN · DAVIS | VFID: 0.159 -> 0.149 | Improved |
| stack 8 layers instead of 6 (ablation) | STTN · DAVIS | SSIM: 95.47 -> 95.60 | Improved |
| use STTN instead of CAP | STTN vs CAP · YouTube-VOS | VFID: 0.071 -> 0.053 | Improved |
| use STTN instead of CAP | STTN vs CAP · YouTube-VOS | PSNR: 31.58 -> 32.34 | Improved |
| use STTN instead of CAP | STTN vs CAP · DAVIS | PSNR: 30.28 -> 30.67 | Improved |
Closest papers by meaning
- Video Inpainting by Jointly Learning Temporal Structure and Spatial Details (AAAI 2019 · similarity 0.92)
- A Temporally-Aware Interpolation Network for Video Frame Inpainting (TPAMI 2019 · similarity 0.90)
- Recurrent Temporal Aggregation Framework for Deep Video Inpainting (TPAMI 2019 · similarity 0.89)
- Internal Video Inpainting by Implicit Long-Range Propagation (ICCV 2021 · similarity 0.89)
- Deep Video Inpainting (CVPR 2019 · similarity 0.89)
- DLFormer: Discrete Latent Transformer for Video Inpainting (CVPR 2022 · similarity 0.88)
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.
