Research map / Video / Video segmentation and tracking
Free-Form Video Inpainting With 3D Gated Convolution and Temporal PatchGAN
Ya-Liang Chang, Zhe Yu Liu, Kuan-Ying Lee, Winston Hsu
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 3D convolution (vs 2D) (ablation) | 3DGated · FVI object-like masks | LPIPS: 0.1769 -> 0.1209 | Improved |
| add 3D convolution (vs 2D) (ablation) | 3DGated · FVI object-like masks | FID: 1.243 -> 1.034 | Improved |
| add gated convolution (vs vanilla) (ablation) | 3DGated · FVI object-like masks | LPIPS: 0.1321 -> 0.1209 | Improved |
| add gated convolution (vs vanilla) (ablation) | 3DGated · FVI object-like masks | FID: 1.121 -> 1.034 | Improved |
| add Temporal PatchGAN loss (ablation) | 3DGated · FVI object-like masks | LPIPS: 0.1716 -> 0.1209 | Improved |
| add Temporal PatchGAN loss (ablation) | 3DGated · FVI object-like masks | FID: 1.201 -> 1.034 | Improved |
| increase channels of vanilla conv instead of gating (ablation) | vanilla conv · FVI | performance | No clear effect |
| 3D gated model vs CombCN (L1-only video inpainter) | 3DGated · FVI curve masks | LPIPS: 0.0794 -> 0.0521 | Improved |
| 3D gated model vs CombCN | 3DGated · FVI curve masks | MSE: 0.0021 -> 0.0024 | Hurt |
| 3D gated model vs EC on bounding-box masks | 3DGated · FaceForensics BBox | LPIPS: 0.0335 -> 0.0395 | Hurt |
| 3D gated model vs patch-based TCCDS | 3DGated · FVI curve masks | FID: 2.105 -> 0.609 | Improved |
| use video inpainting model for 4x super-resolution vs SRGAN | 3DGated · VOR 4x SR | FID: 1.035 -> 1.096 | Hurt |
Closest papers by meaning
- Free-Form Image Inpainting With Gated Convolution (ICCV 2019 · similarity 0.90)
- Deep Learning-Based Image and Video Inpainting: A Survey (IJCV 2024 · similarity 0.88)
- ZeroPatcher: Training-free Sampler for Video Inpainting and Editing (NeurIPS 2025 · similarity 0.88)
- Deep Video Inpainting (CVPR 2019 · similarity 0.88)
- Internal Video Inpainting by Implicit Long-Range Propagation (ICCV 2021 · similarity 0.87)
- An Internal Learning Approach to Video Inpainting (ICCV 2019 · 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.
