Open University seminar — neuromorphic stereo vision and PID
Master's seminar at the Open University of Israel on neuromorphic stereo vision (event-camera correspondence) and a spike-based PID controller, with slides, a demo, and the recorded talk.
This is my master's seminar in Computer Science at the Open University of Israel. It has two parts: neuromorphic stereo vision, and a neuromorphic PID controller.
The vision half is the correspondence problem for event-based stereo: matching pixels across two streams when occlusions, homogeneous regions, and repetitive patterns make local matching ambiguous. The papers use uniqueness and smoothness as global constraints, then a spiking network (coincidence detectors and disparity detectors) to recover depth from disparity, including Osswald et al. (2017) on 3D perception for neuromorphic stereo.
The control half is a spike-based PID: proportional, integral, and derivative terms in the spike domain, pulse-frequency modulation to the motors, and FPGA hardware driving a 4-DoF arm (Jiménez-Fernández et al. on neuromorphic PID).
Slides: https://turgibot.github.io/OPU-Seminar/
Demo: https://www.youtube.com/watch?v=QrYlgrMWYEE
Seminar talk: https://www.youtube.com/watch?v=rCvDy-8IreM
PDF: https://github.com/Turgibot/OPU-Seminar/blob/master/Neuromorphic_Stereo_Vision_and_Robot_Controller%20(3).pdf