Efficient Computer Vision for Embedded Systems
TimeMonday, July 11th10:30am - 12pm PDT
Location3002, Level 3
Event Type
Embedded Systems
DescriptionMany embedded systems are equipped with cameras and can capture vast amounts of visual data continuously. Due to limited energy and wireless data rates, the raw data (i.e., pixels) cannot be transmitted and must be analyzed at the embedded systems. The tutorial provides hands-on experience improving energy efficiency of computer vision. The first speaker will provide step-by-step instruction using PyTorch for quantization, pruning, and knowledge distillation. The second speaker will present the solutions that won the 2020 and 2021 IEEE Low-Power Computer Vision Challenge. The third speaker will explain engineering principles and processes to create reproducible software for machine learning.