Colin Decourt

Colin Decourt

Activity Management Operator

Followers of Colin Decourt439 followers
location of Colin DecourtToulouse, Occitanie, France

Connect with Colin Decourt to Send Message

Connect

Connect with Colin Decourt to Send Message

Connect
  • Timeline

  • About me

    PhD, Machine Learning R&D Engineer @ Exotec

  • Education

    • Université Paul Sabatier Toulouse III

      2020 - 2023
      Doctor of Philosophy - PhD Artificial intelligence

      PhD as part of ANITI (Artificial Natural Intelligence Toulouse Institute). Research topic: Multiple target extraction, identification and tracking for radar using AI.Supervisors: Rufin VanRullen, Thomas Oberlin, Didier SalleThesis manuscript: https://theses.hal.science/tel-04577275v1

    • ENSEIRB-MATMECA

      2017 - 2020
      Engineer's degree Telecommunications engineering / Artificial intelligence

      Activities and Societies: Strong associative implication (student union, INGENIB)

  • Experience

    • SOGETREL

      Jun 2018 - Aug 2018
      Activity Management Operator

      - Establishment and monitoring of technicians' reports following interventions on the Orange network.- Links between customers, technicians and operators.

    • École de technologie supérieure (ÉTS)

      Jun 2019 - Sept 2019
      Research Intern

      Research project about left ventricle segmentation in pediatric MRI for inter-ventricular communication detections. - Generative adversarial networks for left ventricle segmentation in pediatric MRI- Semi-supervised framework was used to reduce number of annotated data for training.- A new weighted cross-entropy loss using distance transform was introduced.- The new loss was used to weight more the points closer to the real boundaries. See: https://doi.org/10.1016/j.compbiomed.2020.103884Project carried out in the Interventional Imaging Lab. (LIVE). Show less

    • NXP Semiconductors

      Feb 2020 - Aug 2020
      Artificial Intelligence and Radar Signal Processing Engineer Intern

      During this internship I worked on simple algorithms for target detection and classification using FMCW radar data. Activities: radar signal processing, prior art about target detection and classification with radar data, development of deep learning models for target classification and detection

    • NXP Semiconductors

      Oct 2020 - Dec 2023
      PhD Student

      - Creation of a deep neural network architecture for object classification using radar data (data collection, model definition, training and evaluation).- Development of a lightweight Faster R-CNN based architecture for range-doppler spectrum features extraction and object detection (adapt model to radar, training, optimization and inference)- Creation of a memory-efficient architecture based on convolutions and convolutional recurrent neural networks for single-view (range-doppler, range-angle) or multi-view (range-azimuth-doppler tensors) object detection and segmentation (model and metric definition, training, optimization and inference).- Development of a self-supervised learning framework for radar object detection using contrastive learning and generative methods (new method definition for radar, implementation, training and inference).- Work on data collection platform for automotive radar applications (system and sensors definition, data collection, data processing).- Accepted articles at IEEE IV 2022 conference and IEEE Transactions in Intelligent Transportation Systems. One other in review. See here for further details: https://colindecourt.github.io/publications/- Assistant professor in machine learning, deep learning, computer vision (classification, detection, segmentation) and computer science (C and Python programming, algorithmic). Details are available here: https://colindecourt.github.io/teaching/- Supervision of two interns (2021, 2022) and one apprentice (2021-2023). Show less

    • ISAE-SUPAERO

      Oct 2020 - Dec 2023
      PhD Student

      - Creation of a deep neural network architecture for object classification using radar data (data collection, model definition, training and evaluation).- Development of a lightweight Faster R-CNN based architecture for range-doppler spectrum features extraction and object detection (adapt model to radar, training, optimization and inference)- Creation of a memory-efficient architecture based on convolutions and convolutional recurrent neural networks for single-view (range-doppler, range-angle) or multi-view (range-azimuth-doppler tensors) object detection and segmentation (model and metric definition, training, optimization and inference).- Development of a self-supervised learning framework for radar object detection using contrastive learning and generative methods (new method definition for radar, implementation, training and inference).- Work on data collection platform for automotive radar applications (system and sensors definition, data collection, data processing).- Accepted articles at IEEE IV 2022 conference and IEEE Transactions in Intelligent Transportation Systems. One other in review. See here for further details: https://colindecourt.github.io/publications/- Assistant professor in machine learning, deep learning, computer vision (classification, detection, segmentation) and computer science (C and Python programming, algorithmic). Details are available here: https://colindecourt.github.io/teaching/- Supervision of two interns (2021, 2022) and one apprentice (2021-2023). Show less

    • Exotec

      Sept 2024 - now
      Machine Learning R&D Engineer
  • Licenses & Certifications