Research map / 3D vision / Depth and stereo
Towards Multimodal Depth Estimation From Light Fields
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace point regression with a discrete softmax posterior | EPI-Net on multimodal light-field validation scenes · Multimodal light-field validation scenes | KL divergence, overall: 5.421 -> 1.734 | Helped |
| Replace point regression with a discrete softmax posterior | EPI-Net on HCI 4D Light Field Dataset · HCI 4D Light Field Dataset | BadPix: 0.065 -> 0.044 | Helped |
| Predict a learned-width Laplacian posterior | EPI-Net on multimodal light-field validation scenes · Multimodal light-field validation scenes | AuSE: UPR (uni) 0.071 | Helped |
| Use shifted-input ensemble inference | EPI-Net on multimodal light-field validation scenes · Multimodal light-field validation scenes | KL divergence, overall: BASE (uni) 5.421; ESE (uni) 4.160 | Hurt |
| Train UPR on all disparity modes instead of only the closest disparity (ablation) | UPR · Multimodal light-field validation scenes | KL divergence, overall: 1.879 -> 2.114 | No clear effect |
| Train UPR on all disparity modes instead of only the closest disparity (ablation) | UPR · Multimodal light-field validation scenes | BadPix: 0.235 -> 0.285 | Hurt |
| Train ESE on all disparity modes instead of only the closest disparity (ablation) | ESE · Multimodal light-field validation scenes | KL divergence, multimodal: 3.628 -> 3.719 | Hurt |
| Train ESE on all disparity modes instead of only the closest disparity (ablation) | ESE · Multimodal light-field validation scenes | BadPix: 0.223 -> 0.229 | Helped |
| Train DPP on all disparity modes instead of only the closest disparity (ablation) | DPP · Multimodal light-field validation scenes | KL divergence, multimodal: 3.057 -> 2.987 | Helped |
| Train DPP on all disparity modes instead of only the closest disparity (ablation) | DPP · Multimodal light-field validation scenes | BadPix: 0.209 -> 0.231 | Hurt |
| Increase ESE's shifted-input ensemble size to cover the large disparity range | ESE · Multimodal light-field validation scenes | runtime: 2.188 -> 17.492 | Hurt |
| Use multimodal rather than single-disparity ground-truth training (ablation) | EPI-Net point-regression baseline · Multimodal light-field validation scenes | BadPix: 0.229 -> 0.307 | Hurt |
Closest papers by meaning
- What Sparse Light Field Coding Reveals About Scene Structure (CVPR 2016 · similarity 0.89)
- EPINET: A Fully-Convolutional Neural Network Using Epipolar Geometry for Depth From Light Field Images (CVPR 2018 · similarity 0.89)
- Attention-Based View Selection Networks for Light-Field Disparity Estimation (AAAI 2020 · similarity 0.88)
- Differentiable Diffusion for Dense Depth Estimation From Multi-View Images (CVPR 2021 · similarity 0.88)
- Behind the Scenes: Density Fields for Single View Reconstruction (CVPR 2023 · similarity 0.88)
- DNGaussian: Optimizing Sparse-View 3D Gaussian Radiance Fields with Global-Local Depth Normalization (CVPR 2024 · 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.