Thiago Rateke

Computer Vision Researcher

Image Processing and Computer Graphics Lab, LAPIX. National Institute for Digital Convergence, INCoD•  August 2009 - August 2019

• Designed an approach for road features semantic segmentation, like surface types (with more than 90% accuracy for each one), potholes (with 97% accuracy), water-puddles (with 90% accuracy), etc. • Developed an approach for road type and quality classification (with more than 90% accuracy in most classes). • Defined an approach for obstacles features extraction, like movement and depth, using CNN, Stereo Vision and Optical Flow (with more than 80% on average accuracy). • Created Ground Truths for road classification (with more than 50000 frames) and for road segmentation (with 701 frames). • Planned and created a new road image dataset, based on emerging countries scenarios, with low-cost cameras. • Helped other project participants to organize and carry out their activities. • Creation and organization of the project website. • Implemented Case-Based Reasoning (CBR) applications (Web, Desktop and Android).

Computer Vision Engineer

Agricultural Research and Rural Extension Company of Santa Catarina, EPAGRI•  June 2015 - December 2015

• Continuity to an on-going project for the visualization and localization of meteorological features. • Fix compilation errors, segmentation faults, update dependencies libraries and small updates on interface. • Build a script for installation considering dependencies installations.

Developer

Brazillian Health Regulatory Agency, ANVISA•  June 2013 - June 2014

• Improved and organized the evaluations and data collect for Anvisa inspectors, designing and developing an android app. • Developed websites so that establishments could register and self-assess themselves, as well as an administrative area for Anvisa inspectors and coordinators.

Intern

Númera Solutions and Systems•  March 2008 - August 2008

• Tests on company systems and websites development.

Universidade Federal de Santa Catarina

Graduate Program in Computer Science (PPGCC), PhD•  August 2015 - May 2020

Road detection adaptive to surface type and detection of road surface damages for vehicle navigation tasks using visual perception and Machine Learning methods.

Universidade Federal de Santa Catarina

Graduate Program in Computer Science (PPGCC), MS•  March 2012 - March 2015

In the final project of my master degree I developed a methodology to control the trajectory of a small vehicle using techniques from Computer Vision and Digital Image Processing.

Universidade do Vale do Itajaí

Computer Science, BS•  March 2007 - December 2011

In my bachelor degree final project I used image segmentation and pixel correlation methods (Stereo Vision) to divide the image into subregions and quantify disparity on the segmented images.