Research map / 3D vision / Depth and stereo
End-to-end View Synthesis for Light Field Imaging with Pseudo 4DCNN
Yunlong Wang, Fei Liu, Zilei Wang, Guangqi Hou, Zhenan Sun, Tieniu Tan
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 prior sensitive loss (ablation) | Pseudo 4DCNN · 30 Scenes | PSNR (dB): 42.34 -> 43.28 | Improved |
| use a separately trained row and column network instead of the same sub-network (Fr=Fc) (ablation) | Pseudo 4DCNN · 30 Scenes | PSNR (dB): 42.90 -> 43.28 | Improved |
| add 2D learnable interpolation (ablation) | Pseudo 4DCNN · 30 Scenes | PSNR (dB): 40.15 -> 43.28 | Improved |
| add 3D detail-restoration CNN (ablation) | Pseudo 4DCNN · 30 Scenes | PSNR (dB): 39.01 -> 43.28 | Improved |
| add prior sensitive loss (ablation) | Pseudo 4DCNN · 30 Scenes | SSIM: 0.9901 -> 0.9916 | Improved |
| add 3D detail-restoration CNN (ablation) | Pseudo 4DCNN · 30 Scenes | SSIM: 0.9876 -> 0.9916 | Improved |
| replace Wu et al. EPI blur-restoration-deblur with Pseudo 4DCNN | Wu et al. · 30 Scenes | PSNR (dB): 40.83 -> 43.28 | Improved |
| replace Kalantari et al. disparity-based synthesis with Pseudo 4DCNN | Kalantari et al. · 30 Scenes | PSNR (dB): 37.43 -> 43.28 | Improved |
| replace Wu et al. with Pseudo 4DCNN | Wu et al. · 30 Scenes | time per synthesized view (sec): 3.38 -> 0.28 | Improved |
| replace Kalantari et al. with Pseudo 4DCNN | Kalantari et al. · 30 Scenes | time per synthesized view (sec): 8.59 -> 0.28 | Improved |
| replace Wu et al. with Pseudo 4DCNN on non-Lambertian synthetic scenes | Wu et al. · Kitchen | PSNR (dB): 35.57 -> 38.12 | Improved |
| replace Wu et al. with Pseudo 4DCNN on iris light fields | Wu et al. · Iris | PSNR (dB): 34.98 -> 40.14 | Improved |
Closest papers by meaning
- Learning to Synthesize a 4D RGBD Light Field From a Single Image (ICCV 2017 · similarity 0.90)
- A Generative Model for Generic Light Field Reconstruction (TPAMI 2020 · similarity 0.88)
- Learning Light Field Angular Super-Resolution via a Geometry-Aware Network (AAAI 2020 · similarity 0.88)
- High-Dimensional Dense Residual Convolutional Neural Network for Light Field Reconstruction (TPAMI 2019 · similarity 0.88)
- Residual Networks for Light Field Image Super-Resolution (CVPR 2019 · similarity 0.88)
- Fast Light Field Reconstruction With Deep Coarse-To-Fine Modeling of Spatial-Angular Clues (ECCV 2018 · 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.
