Non-Line-of-Sight Imaging from iToF data

The master's thesis will be about recovering information regarding an object out of line of sight. In the target set-up, multi-frequency images acquired by an iToF camera looking at an intermediate wall are used in combination with a neural network for direct-global light separation. The first part consists on, using analytical approaches, similar to "Fermat Flow", to estimate the location of the object around the corner. While the second part consists in using a Deep Learning model to perform the same task. The method will be tested on synthetic scene, simulated with Mitsuba-v2, and real scenes.