Machine Learning Enhanced Compressive Hyperspectral Imaging - IPAM at UCLA

Explore machine learning-enhanced compressive hyperspectral imaging techniques in this conference talk presented by Kevin Kelly at IPAM's Multi-Modal Imaging Workshop. Delve into approaches combining compressive imaging systems and neural network algorithms for hyperspectral machine vision tasks, implemented in reconstruction and directly on compressive measurements. Discover how spatial light modulators perform optical computation in convolutional neural networks' first layer and their application in object recognition using neural networks as nonlinear transforms. Learn about a dynamic sampling rate approach for training neural networks specifically for compressive measurements in optical systems. Examine the L1 compressive foveation result and its applications in parallelizing large image reconstruction and replacing optimization algorithms with neural networks for rapid reconstruction. Gain insights into various topics, including single-pixel CS cameras, CS imaging in infrared, dark-field microscopy, micro-extinction spectroscopy, compressive hyperspectral microscopy systems, and compressed domain classification.

Machine Learning Enhanced Compressive Hyperspectral Imaging
"Single-Pixel" CS Camera
CS Imaging in the Infrared
Dark-field Microscopy
Micro-Extinction Spectroscopy (MEXS) Setup
Compressive Hyperspectral Microscopy System
CS Endmember Unmixing
CS Machine Vision
Compressive Matched Filtere
Convolutional Neural Network
Hybrid Optical Compressed CNN
Hardware HOC-CNN
Dynamic-Rate Neural Network ce
Compressed Domain Classification
Compressed Sensing Machine Vision
CS Regional Foveation
Foveated Parallel Reconstruction
Compressive Sensing Software