Davi Frossard

Computer Vision Researcher

About me


I'm a graduate student at the University of Toronto doing research on computer vision under the supervision of prof. Raquel Urtasun. My research interests lie in the intersection of classic computational geometry, vision and machine learning.

Previously, I did my Master's at the University of Toronto with a thesis on extracting visual vehicle attributes from a self driving platform. I did my Bachelor's degree at the Federal University of Espirito Santo (UFES) in Computer Engineering, writing my thesis on end-to-end learning of multiple object tracking.

Interests

  • Computer Vision

  • Computational Geometry

  • Deep Learning

Education

  • PhD in Computer Science

    University of Toronto, 2023

  • MSc in Computer Science

    University of Toronto, 2018

  • BSc in Computer Engineering

    Federal University of Espirito Santo, 2017

Publications

StrObe: Streaming Object Detection from LiDAR Packets


Davi Frossard, Simon Suo, Sergio Casas, James Tu, Rui Hu, Raquel Urtasun

Existing LiDAR perception systems wait 100ms just to build a full sweep. StrObe instead does streaming detection from LiDAR packets and uses a spatial memory to leverage temporal information, achieving an end-to-end latency of 21ms (a 6x speedup!) and outperforming the state of the art in detection.

DeepSignals: Predicting Intent of Drivers Through Visual Signals


Davi Frossard, Eric Kee, Raquel Urtasun

Turn signals and emergency flashers communicate intention of drivers for sudden events like lane changes and stops. Recognizing these signals provides valuable foresign to autonomous vehicles. In this paper, we propose to detect these signals by using a deep neural network that reasons about both spatial and temporal information.

End-To-End Learning of Multi-Sensor 3D Tracking by Detection


Davi Frossard, Raquel Urtasun

In this paper we propose a novel approach to tracking by detection by formulating the problem as a linear program that can be solved exactly, and learn convolutional networks for detection as well as matching in an end-to-end manner.

Copyright © 2021 Davi Frossard.