Research map / 3D vision / Depth and stereo
EPINET: A Fully-Convolutional Neural Network Using Epipolar Geometry for Depth From Light Field Images
Read the paperCheck 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 |
|---|---|---|---|
| Use four geometry-specific streams instead of a one-stream network, with the same number of parameters (ablation) | EPINET · Light-field depth estimation | reconstruction error: about 10 percent lower | Helped |
| Increase the input angular views from 7×7 to 9×9 (ablation) | EPINET · Synthetic light-field data | mean square error | Hurt |
| Increase the input angular views from 7×7 to 9×9 (ablation) | EPINET · Real-world light-field images | disparity estimation performance | Helped |
| Add light-field-specific augmentation, including view shifting, rotation, flipping, and scaling (ablation) | EPINET · Light-field depth estimation | disparity error: more than 40% | Helped |
| Add rotation and flipping augmentation (ablation) | EPINET · Light-field depth estimation | performance | Helped |
| Use 9×9 rather than 7×7 input views (ablation) | EPINET · Synthetic light-field data | performance | No clear effect |
Closest papers by meaning
- Light Field Reconstruction Using Deep Convolutional Network on EPI (CVPR 2017 · similarity 0.90)
- Light Field Intrinsics With a Deep Encoder-Decoder Network (CVPR 2018 · similarity 0.90)
- Convolutional Networks for Shape From Light Field (CVPR 2016 · similarity 0.89)
- Towards Multimodal Depth Estimation From Light Fields (CVPR 2022 · similarity 0.89)
- Residual Networks for Light Field Image Super-Resolution (CVPR 2019 · similarity 0.89)
- What Sparse Light Field Coding Reveals About Scene Structure (CVPR 2016 · similarity 0.88)
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.