Research map / 3D vision / Depth and stereo
Learning to See Through Obstructions
Yu-Lun Liu, Wei-Sheng Lai, Ming-Hsuan Yang, Yung-Yu Chuang, Jia-Bin Huang
Check 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 the proposed multi-frame flow decomposition and reconstruction method | reflection removal · Synthetic sequences, background and reflection layers | PSNR, SSIM, NCC, and LMSE: Background: PSNR 23.62 -> 26.57, SSIM 0.8192 -> 0.8676, NCC 0.9023 -> 0.9380, LMSE 0.0177 -> 0.0125; reflection: PSNR 21 | Improved |
| Add uniform flow initialization instead of zero initialization (ablation) | the proposed method · Vimeo-90k validation | validation loss: 0.377 -> 0.184 | Improved |
| Add uniform flow initialization instead of dense flow initialization (ablation) | the proposed method · Vimeo-90k validation | validation loss: 0.226 -> 0.184 | Improved |
| Add the image reconstruction network instead of temporal mean filtering (ablation) | the proposed method · Vimeo-90k validation | validation loss: 0.526 -> 0.184 | Improved |
| Add the image reconstruction network instead of temporal median filtering (ablation) | the proposed method · Vimeo-90k validation | validation loss: 0.482 -> 0.184 | Improved |
| Add synthetic-data pre-training (ablation) | the proposed method with online optimization · Vimeo-90k validation | validation loss: 0.417 -> 0.139 | Improved |
| Add online optimization (ablation) | the proposed method with pre-training · Vimeo-90k validation | validation loss: 0.184 -> 0.139 | Improved |
Closest papers by meaning
- Learning to See Through Obstructions With Layered Decomposition (TPAMI 2021 · similarity 0.99)
- Instruct2See: Learning to Remove Any Obstructions Across Distributions (ICML 2025 · similarity 0.88)
- Seeing Deeply and Bidirectionally: A Deep Learning Approach for Single Image Reflection Removal (ECCV 2018 · similarity 0.87)
- A Generic Deep Architecture for Single Image Reflection Removal and Image Smoothing (ICCV 2017 · similarity 0.87)
- Neural Spline Fields for Burst Image Fusion and Layer Separation (CVPR 2024 · similarity 0.85)
- Gated Fields: Learning Scene Reconstruction from Gated Videos (CVPR 2024 · similarity 0.85)
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.
