Get Started
Home
Topics
Search
Library
Research questionHow can mirror inpainting generate geometry-consistent reflections from a fixed scene without hallucinating visible content?A mirror region is not an ordinary missing image area: its appearance depends on scene geometry and viewpoint, while generative models may invent content already visible elsewhere.
AI
Computer Vision
Diffusion Models
Evaluation & Benchmarks
Image & Video Processing
Image Generation
Latest papersRecent research connected to this question, newest first.Fill My Mirror: Geometry-Constrained Mirror InpaintingThe source addresses fixed-scene mirror inpainting using estimated geometry, projected visible content, and a training-free two-mask diffusion process. Evidence covers complex real-world scenes, synthetic MirrorBench-V2 images, real images, and standard plus geometry-aware consistency metrics.research paper · Sep 3, 2026
Related questions
How can novel-view synthesis use mirror reflections to generate consistent views and recover hidden scene structure?How can generalized scene reconstruction preserve fine geometric detail for accurate rendering and physical measurement?How can vision detect water-reflection axes to reduce false detections in natural-water scenes?How can we complete unseen 3D scene regions from sparse, unconstrained views without dense 3D supervision?