Reconstructing Unseen Dimensions in Biophotonics Through Python
summary
Combines a high‑resolution SIMOS intensity volume (up to ~2000×2000 with ~20 z‑slices) and a low‑res single‑pixel camera tensor (32×32 spatial with spectral and lifetime channels) into a reconstructed 5D high-resolution tensor (x,y,z,λ,t) by solving a convex data‑fusion loss with total‑variation regularization. Uses DMD/Hadamard patterns, spectrometer and TCSPC, implements operators and their transposes in Python/PyTorch on MPS, solves with conjugate gradient in PyTorch (converges ~17–18 iterations), and validates against hardware ground truth with PSNR/SSIM.
AI-written summary · about
chapters (7)
- 0:00Introduction and background
- 2:14Problem in biophotonics
- 9:50Single pixel camera basics
- 14:08Data fusion approach
- 17:23Python implementation
- 20:01Results and validation
- 22:12Q&A session
related talks
10mHow DeepInverse Is Solving Imaging in Science and H...
01h 14mRecent Progress in High-Dimensional Learning
45mSolving Inverse Problems with Deep Learning
49mGraph Based Processing of Big Images
17mData Challenges with 3D Computer Vision #bbuzz
42mLightFields.jl: Fast 3D image reconstruction for VR applications
09mLightning Talk: A Novel Domain Generalization Technique for Medical Imaging Using...
26mAccelerating Explorations in Vision and Multimodal AI Using Pytorch...