Research map / Video / Video segmentation and tracking
Short-Term and Long-Term Context Aggregation Network for Video Inpainting
Ang Li, Shanshan Zhao, Xingjun Ma, Mingming Gong, Jianzhong Qi, Rui Zhang, Dacheng Tao, Ramamohanarao Kotagiri
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 |
|---|---|---|---|
| add BSCA (multi-scale boundary-aware short-term aggregation) (ablation) | STCA-Net · DAVIS (shuffled object masks, ablation) | PSNR: 27.17 -> 28.13 | Improved |
| add BSCA (multi-scale boundary-aware short-term aggregation) (ablation) | STCA-Net · DAVIS (shuffled object masks, ablation) | VFID: 0.063 -> 0.042 | Improved |
| apply BSCA at all three encoding scales instead of only 1/8 (ablation) | STCA-Net · DAVIS (shuffled object masks, ablation) | PSNR: 27.72 -> 28.13 | Improved |
| apply BSCA at all three encoding scales instead of only 1/8 (ablation) | STCA-Net · DAVIS (shuffled object masks, ablation) | VFID: 0.052 -> 0.042 | Improved |
| add DLCA (dynamic long-term context aggregation) (ablation) | STCA-Net · DAVIS (shuffled object masks, ablation) | PSNR: 27.58 -> 28.13 | Improved |
| add DLCA (dynamic long-term context aggregation) (ablation) | STCA-Net · DAVIS (shuffled object masks, ablation) | VFID: 0.056 -> 0.042 | Improved |
| use dynamically updated long-term features instead of fixed sampling (every 5th frame) (ablation) | STCA-Net · DAVIS (shuffled object masks, ablation) | PSNR: 27.85 -> 28.13 | Improved |
| use dynamically updated long-term features instead of fixed sampling (every 5th frame) (ablation) | STCA-Net · DAVIS (shuffled object masks, ablation) | VFID: 0.048 -> 0.042 | Improved |
| add fixed-sampling DLCA to BSCA-only model (ablation) | STCA-Net · DAVIS (shuffled object masks, ablation) | PSNR: 27.58 -> 27.85 | Improved |
| set long-term feature set length q to 10 (ablation) | STCA-Net · DAVIS | inpainting quality | Improved |
| compared with FGNet, STCA-Net is faster at similar or better quality | STCA-Net vs FGNet · YouTube-VOS | execution time (sec): 276.3 -> 35.4 | Improved |
Closest papers by meaning
- Recurrent Temporal Aggregation Framework for Deep Video Inpainting (TPAMI 2019 · similarity 0.91)
- Internal Video Inpainting by Implicit Long-Range Propagation (ICCV 2021 · similarity 0.88)
- Contextual Residual Aggregation for Ultra High-Resolution Image Inpainting (CVPR 2020 · similarity 0.88)
- Video Inpainting by Jointly Learning Temporal Structure and Spatial Details (AAAI 2019 · similarity 0.88)
- Semantic-Aware Dynamic Parameter for Video Inpainting Transformer (ICCV 2023 · similarity 0.88)
- Learning Joint Spatial-Temporal Transformations for Video Inpainting (ECCV 2020 · 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.
