@ INF3710 - Fichiers et base de données: Répétiteur et chargés de travaux pratiques

@ DSG (Data Systems Group) under the supervision of Professor Amine Mhedhbi.

◦ Collaborated on the development of a Text-to-3D Avatar Generation pipeline.◦ Fine-tuned a Generative Diffusion Model to output high quality faces from text descriptions.◦ Spearheaded efforts to build a pipeline that generates a dataset of high quality 2D faces and their descriptions using a visual large language model (LLM).◦ Implemented a new evaluation metric to measure the quality of prompt-driven 3D avatar generation and used it to compare our trained model to prior work.

• Implemented and evaluated multiple deep learning model architectures for the semantic segmentation of geospatial high resolution remote sensing imagery using PyTorch.• Researched and compared State of the Art and novel model architectures in semantic segmentation and feature extraction.• Conducted experiments to evaluate the effect of new loss combinations on model training.• Hands-on experience in working with Geographic Information Systems (GIS) to visualize and analyze remote sensing data.

• Collaborated on the development of a remotely operated vehicle (ROV).• Designed an Object Oriented desktop application with a Python backend controlling an ROV through ROS-based communication with STM microcontrollers.• Developed an algorithm using OpenCV and classical computer vision techniques to allow the ROV to follow a line autonomously underwater.