Detecting people using depth frames and deep classifiers

Apr 6, 2017 18:00
DataScience Floripa Meetup - Florianópolis, Brazil

A short talk about a project I took part during my internship at Instituto SENAI de Tecnologia and which was also my final year project for my undergraduate program. My contribution was to create an automatic human detection system for complex industrial environments. We used depth images from a overhead camera position to perform detection.

Two approaches to the detection problem were implemented and evaluated: traditional learning algorithms with manual feature extraction and deep learning classifiers. During this talk I revealed details and tradeoffs of both methods as well as a quantitative evaluation of their performance.

Eduardo Arnold
Eduardo Arnold
Machine Learning Engineer

I’m a Machine Learning Engineer at Niantic. Previously, I obtained my PhD degree at the University of Warwick, supervised by Mehrdad Dianati and Paul Jennings, and focusing on perception methods for autonomous driving.