Research map / Video / Video segmentation and tracking
ProPainter: Improving Propagation and Transformer for Video Inpainting
Shangchen Zhou, Chongyi Li, Kelvin C.K. Chan, Chen Change Loy
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 dual-domain propagation with mask-guided sparse Transformer (full ProPainter vs E2FGVI) | ProPainter · DAVIS | PSNR: 33.01 -> 34.47 | Improved |
| add global image propagation (ablation) | ProPainter · YouTube-VOS ablation | PSNR: 33.05 -> 34.15 | Improved |
| use reliability-checked image propagation instead of FGVC's (ablation) | ProPainter · YouTube-VOS ablation | PSNR: 32.91 -> 34.15 | Improved |
| add feature propagation (ablation) | ProPainter · YouTube-VOS ablation | PSNR: 33.17 -> 34.15 | Improved |
| use own feature propagation instead of E2FGVI's (ablation) | ProPainter · YouTube-VOS ablation | PSNR: 33.94 -> 34.15 | Improved |
| use mask-guided sparse Transformer instead of full-token self-attention (ablation) | ProPainter · YouTube-VOS ablation | PSNR: 34.18 -> 34.15 | No clear effect |
| use recurrent flow completion network instead of prior flow completion | ProPainter · YouTube-VOS flow completion | runtime (s/frame): 0.312 -> 0.005 (vs FGT), EPE 0.020 | Improved |
| apply ProPainter to YouTube-VOS rather than DAVIS | ProPainter · YouTube-VOS vs DAVIS | PSNR gain over E2FGVI: 33.71 -> 34.43 (YouTube-VOS) vs 33.01 -> 34.47 (DAVIS) | Mixed |
Later papers that compared with this method
20 later accepted papers report ProPainter as a baseline in their results tables; 16 of them report a better number on most of the results they share with it.
- CLEAR: Context-Aware Learning with End-to-End Mask-Free Inference for Adaptive Video Subtitle Removal (ICML 2026 · better on 5 of 6 shared results)
- From Draft to Draft-Free: One-Step Video Object Removal via Privileged Distillation and Fast Planting (ECCV 2026 · better on 5 of 6 shared results)
- Generating Humanless Environment Walkthroughs from Egocentric Walking Tour Videos (CVPR 2026 · better on 5 of 5 shared results)
- YOSE: You Only Select Essential Tokens for Efficient DiT-based Video Object Removal (CVPR 2026 · better on 0 of 4 shared results)
- DreamStereo: Towards Real-Time Stereo Inpainting for HD Videos (CVPR 2026 · better on 4 of 4 shared results)
- Object-WIPER: Training-Free Object and Associated Effect Removal in Videos (CVPR 2026 · better on 2 of 4 shared results)
- S$^{2}$VG: 3D Stereoscopic and Spatial Video Generation via Denoising Frame Matrix (TPAMI 2026 · better on 4 of 4 shared results)
- Trajectory-Stabilized Inference for Diffusion-Based Video Inpainting (ICML 2026 · better on 1 of 3 shared results)
12 more not listed.
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Video Diffusion Models Are Strong Video Inpainter (AAAI 2025 · similarity 0.89)
- WaveFormer: Wavelet Transformer for Noise-Robust Video Inpainting (AAAI 2024 · similarity 0.87)
- Keyframe-Guided Creative Video Inpainting (CVPR 2025 · similarity 0.86)
- MotionAura: Generating High-Quality and Motion Consistent Videos using Discrete Diffusion (ICLR 2025 · similarity 0.86)
- Elevating Flow-Guided Video Inpainting with Reference Generation (AAAI 2025 · similarity 0.86)
- Reconstruct, Inpaint, Test-Time Finetune: Dynamic Novel-view Synthesis from Monocular Videos (NeurIPS 2025 · similarity 0.86)
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.
