R&D Project in collaboration with The French Federation of Tennis- AI Solution for Tennis Analysis with the National Technical Direction of Tennis in France.- Novel Monocamera Approach: Minimizing Multi-Camera and Sensor Dependency for Advanced Tennis Performance Analysis.- Enhancing Player Performance: Leveraging Deep Learning for Training and In-Game Analysis.

Design & Creation of a Video analysis system of player strategy in the game of badminton with deep learning- Survey in visual objects detection and tracking of player / shuttlecock / court- Implementation of an action recognition 3DCNN-Based model and Classification of each player's strategy.- Built full dataset of skeleton data describing different strokes in the game of badminton.- Proposed new way of generating statistics to understand each player's performance :… Show more Design & Creation of a Video analysis system of player strategy in the game of badminton with deep learning- Survey in visual objects detection and tracking of player / shuttlecock / court- Implementation of an action recognition 3DCNN-Based model and Classification of each player's strategy.- Built full dataset of skeleton data describing different strokes in the game of badminton.- Proposed new way of generating statistics to understand each player's performance : - Aerial Hit-maps of player's positions evolution - Style of play : offensive / defensive Show less

NLP Flowchart: from input text to statistics and visualiation- Created custom scraper to retrieve german language news articles from major news websites.- Explained & evaluated embeddings models: Shallow BoW like doc2vec & TF-IDF, and deep pretrained Single Vector Encoder: BERT on our data.- Topic modeling visualization as part of products for the clients.

Image Classification model :- Data preprocessing and cleaning- Fine-tuned a pretrained VGG16 model on our data.- Proposed to integrate an attention layer with VGG-16 for more efficient object recognition. Based on the paper: The Costs and Benefits of Goal-Directed Attention in Deep Convolutional Neural Networks, Luo, Xiaoliang and Roads, Brett and Love, Bradley.- Reached 92 % de précision top-1 et 97,4 % de précision top-5.

Chatbot Development for E-Learning Platform:- Developed an intelligent chatbot leveraging Microsoft Azure's LUIS.ai, as the core Natural Language Processing (NLP) model, to perform advanced Natural Language Processing (NLP) tasks, including assisting users with searching for courses, books, and workshops in an e-learning library.- Implemented a scalable chatbot architecture, ensuring seamless interaction with users and robust integration with backend systems.- Integrated Microsoft… Show more Chatbot Development for E-Learning Platform:- Developed an intelligent chatbot leveraging Microsoft Azure's LUIS.ai, as the core Natural Language Processing (NLP) model, to perform advanced Natural Language Processing (NLP) tasks, including assisting users with searching for courses, books, and workshops in an e-learning library.- Implemented a scalable chatbot architecture, ensuring seamless interaction with users and robust integration with backend systems.- Integrated Microsoft Graph API to collect model predictions and deliver accurate and context-aware responses to user queries.- Trained the LUIS.ai model by providing domain-specific prompts of user messages, mapping intents and entities to optimize the model's accuracy and relevance for the platform's needs. Show less