Autonomous cars are close to become a reality. But are we prepared to have autonomous cars in all kinds of scenarios? Deep Neural Networks are extensively used on Video and Lidar sensors showing to have outstanding results for tasks like object recognition and mapping. However those benchmark results are in good weather and those sensors perform poorly in bad weather conditions. Radar sensors are robust to rain, fog and snow. Infrared sensors are capable to sense in low lighting conditions. However radar and infrared datasets are quite limited and have lower resolution compared to video. My project aims to transfer the knowledge from big video datasets to radar and infrared sensors in order to achieve more reliable object recognition in bad weather scenarios. The main research topics of my project are Deep Neural Networks, Transfer Learning and Domain Adaptation.
- 2016 - Present - PhD in Computer Vision and Robotics, Heriot-Watt University, Edinburgh, UK
- PhD Project: Multi-Modal Transfer Learning for Object Recognition applied to Autonomous Cars in Bad Weather
- 2016 - Summer Internship at Codeplay Software, Edinburgh, UK
- C++ Computer Vision Engineer Intern
- 2014 - 2016 - MSc in Computer Vision and Robotics, Heriot-Watt University, Universitat de Girona and Université de Bourgogne
- MSc Thesis title: Deep Convolutional Poses for Human Interaction Recognition
- 2014 - Summer Internship at The University of British Columbia, Vancouver, Canada
- Project: Automated image analysis for measurement of brain structures on magnetic resonance images.
- Supervised by Roger Tam.
- 2011 - 2014 BSc in Computer Science - Universidade Federal Fluminense, Rio de Janeiro, Brazil
- Exchange period at Aarhus Universitet, Denmark
- BSc Thesis title: Machine Learning Methods applied to Breast Diseases using Infrared Images