Research map / 3D vision / Depth and stereo
FlatNet: Towards Photorealistic Scene Reconstruction from Lensless Measurements
Salman Siddique Khan, Varun Sundar, Vivek Boominathan, Ashok Veeraraghavan, Kaushik Mitra
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 calibrated (transpose) initialization instead of random (ablation) | FlatNet-sep · Display captured FlatCam | PSNR (dB): 19.06 -> 19.62 | Improved |
| add calibrated PSF initialization instead of simulated (ablation) | FlatNet-gen · Display captured PhlatCam | PSNR (dB): 20.53 -> 20.94 | Improved |
| add calibrated PSF initialization (ablation) | Le-ADMM · Display captured PhlatCam | PSNR (dB): 18.35 -> 20.29 | Improved |
| add trainable inversion stage in place of Tikhonov + U-Net (ablation) | FlatNet-sep · Display captured FlatCam | PSNR (dB): 18.90 -> 19.62 | Improved |
| add trainable inversion stage in place of Tikhonov + U-Net (ablation) | FlatNet-gen · Display captured PhlatCam | PSNR (dB): 20.60 -> 20.94 | Improved |
| add trainable inversion stage in place of Tikhonov + U-Net (ablation) | FlatNet-sep · Display captured FlatCam | LPIPS: 0.322 -> 0.256 | Improved |
| use FlatNet-gen instead of Le-ADMM on cropped measurements | FlatNet-gen-C · Cropped display captured PhlatCam | PSNR (dB): 18.72 -> 19.29 | Improved |
| use uncalibrated PSF on cropped measurements (ablation) | Tikh+U-Net · Cropped display captured PhlatCam | PSNR (dB): 18.34 -> 17.53 | Hurt |
| use FlatNet-gen instead of Le-ADMM (memory) | trainable inversion · 1280x1408 measurement | memory (MB): 6300 -> 990 | Improved |
| use FlatNet-gen instead of TV-ADMM for speed | FlatNet-gen · Display captured PhlatCam | inference time (s): 180 -> 0.03 | Improved |
Closest papers by meaning
- Towards Photorealistic Reconstruction of Highly Multiplexed Lensless Images (ICCV 2019 · similarity 0.94)
- Integrated Forward–Inverse Network for Reconstruction for Lensless Image Reconstruction (ECCV 2026 · similarity 0.89)
- A Simple Framework for 3D Lensless Imaging With Programmable Masks (ICCV 2021 · similarity 0.88)
- PhoCoLens: Photorealistic and Consistent Reconstruction in Lensless Imaging (NeurIPS 2024 · similarity 0.87)
- 3D Gaussian Flats: Hybrid 2D/3D Photometric Scene Reconstruction (NeurIPS 2025 · similarity 0.86)
- TurboSL: Dense Accurate and Fast 3D by Neural Inverse Structured Light (CVPR 2024 · similarity 0.84)
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.
