Physics-Informed Imaging Systems

This lecture by the instructor covers the development of cameras optimized for machine learning tasks and the use of machine learning to enhance camera capabilities. Topics include physics-informed machine learning for lensless imaging, deep learning for imaging inverse problems, and the application of algorithms to solve imaging challenges. The lecture explores the optimization of cameras for high-resolution images, the use of deep learning for spatially-varying microscopy, and the training of denoisers using synthetic noisy videos. Additionally, it delves into the creation of noise models for low-light videos, the simulation of supervised denoising, and the advancement of computational cameras through physics-informed machine learning. The presentation showcases the instructor's work on pushing the limits of cameras with GAN-tuned noise models and the achievement of photorealistic videography in low-light conditions.