M2 internship: March-August 2021 (6 months)
Quantification of deformation on right ventricular mesh sequences
CREATIS laboratory (CNRS UMR 5220, INSERM U1294, Université Lyon 1, INSA Lyon, Université Jean Monnet Saint-Etienne) 🇫🇷
- Evaluating RV Quantification: Currently there is no consensus on right ventricular quantification. Our goal is to implement state of the art quantification method for shape and strain; That will allow us to evaluate each methods’ impact and stability under noise.
- Perspectives: The work done during this internship could be used to make a consensus paper between researchers and the industry by my tutors, Dr. Nicolas Duchateau and Dr. Gabriel Bernardino.
M1 internship: June-August 2020 (3 months)
Automatic segmentation of weakly supervised medical images
LIVIA laboratory, ÉTS Montréal, Canada 🇨🇦
- Deep learning with geometrically constrained losses: Based on the previous work of Pr. Ismail Ben Ayed and Dr. Hoel Kervadec, we extended the use of constraints on bounding boxes to 3D. Their previous works used geometrical constraints enforced in the loss of a CNN to segment organs that were weakly supervised (only a portion of the pixels labeled, bounding boxes...). We tested if their results still worked when the bounding box was defined in 3D, and for multi-organ segmentation.
- We submitted a publication to Medical Image Analysis
- UNIQUE-Mitacs Globalink Research Grant
Youth worker/coordinator 2019-2020 (1 years)
“Francas de Seine Maritime”, St-Étienne-du-Rouvray, France 🇫🇷
- Following my civic service, I was recruited by the organization to lead the technical team projects for a year.
- Conducting "hacker spaces" type workshops for pre-teens and teenagers.
- Technical team coordinator (september 2019 - june 2020): Management of a team of 3 to 6 people, internship mentoring (3 interns for 2 months), development of many projects (humanoid robot, ESP, Drones, 3D printers...)
L3 internship: April-July 2019 (3 months)
LITIS laboratory EA 4108, University of Rouen Normandy 🇫🇷
- Training an OCR on a handwritten dataset: Following my L2 internship, I used a homemade library (OCR CTC, by Dr. Yann Soullard) based on Keras to train an OCR like the one I used to get results from.
- Evaluating the capacity of adaptation of a deep learning model to specific writers: We wondered if a model could be trained to improve its result on a specific writer, with few examples. The results were inconclusive. We used CNN and BLSTM, L2-SP regularization, Data augmentation (Elastic distortion). - thomasdargent.com/stage-l3
Youth worker 2018-2019 (1 year)
“Francas de Seine Maritime” organization, St-Étienne-du-Rouvray, France 🇫🇷
- Child Counseling and Popular Education : Teaching weekly technology workshop for childrens (8 to 15 years old), 4 of our teams did the Eurobot Junior, a robotic competition for teenagers.
- Research and Development: InMoov, 3D printers, A syste for children to control a humanoid robot (Named OEHBI later).
L2 internship: June-July 2018 (2 months)
Joint laboratory between LITIS (EA 4108) and the company HAMELIN
- Voluntary internship: The project was to have a web based administration system for an OCR (correcting it's mistakes, seeing the output...). I had to work on the logging system.
- Added a logging system and developed administration features: ELK, VueJS & Django
Summer Job August 2016
Antenne Normandie 🇫🇷
- Setting Up a Network of Security Cameras