Ayush Jamdar (UCSD): “ISET-LFM: A Physics-based Simulation Framework and Dataset for LED Flicker in Automotive Imaging”
Speaker: Ayush Jamdar (UCSD)
Title: ISET-LFM: A Physics-based Simulation Framework and Dataset for LED Flicker in Automotive Imaging
Video:
Abstract: LED flicker is a persistent artifact in automotive imaging where lights modulated via Pulse Width Modulation (PWM) appear steady to humans but produce severe temporal intensity variations in captured video. These artifacts can lead to the misidentification of traffic signals, turn and brake signals, or emergency strobes in Advanced Driver Assistance Systems (ADAS), posing significant known safety risks. While hardware mitigations like split-pixel sensor architectures reduce flicker, they introduce a fundamental trade-off with motion blur. This work explores the development of ISET-LFM, an open-source physics-based simulation framework built upon the ISET ecosystem, specifically ISET3D and ISETCam, to model the entire imaging chain from 3D scene dynamics to detailed sensor electronics. We will discuss the integration of an analytical PWM flicker model with active camera transforms to simulate realistic non-uniform motion blur. The framework enables the generation of simultaneous dual-exposure radiance maps alongside guaranteed flicker-free ground truth, providing a critical resource for benchmarking and training learned LED flicker mitigation (LFM) algorithms.
Bio: Ayush M. Jamdar is a graduate student at the University of California San Diego, where he is specializing in Signal and Image Processing within the Electrical and Computer Engineering department. He is a member of the UCSD Computational Imaging Systems Lab (CISL) and holds a B.Tech in Electrical Engineering from the Indian Institute of Technology Madras. During the summer of 2025, Ayush served as a Camera Systems Intern at Omnivision Technologies and a Research Affiliate at Stanford University’s VISTA Lab. It was during this time that he developed the ISET-LFM simulation framework and dataset to address the data bottleneck in automotive LFM research. His professional interests lie at the intersection of computational imaging, physically-based simulation, and machine learning.
