Projects
You can find here a list my most relevant projects.
Literature Review : Neural Audio Editing Techniques
Neural Audio Editing TechniquesDescription: This report presents a brief literature review of recent neural audio editing techniques and highlights relevant image editing approaches that have not yet been explored in the audio domain. It concludes with intuitions on research directions for potential improvements in efficiency and performance.
Keywords: Audio Editing, Image Editing, Literature review, Diffusion Models, Inversion techniques
Exploring Audio AI
Exploring Audio AIDescription: This document has a dual purpose. First, it aims to provide a concise overview of how Artificial Intelligence (AI) is recently evolving in the field of audio and music. Additionally, it intends to explore potential research directions based on the current state of the domain and my own perspectives and interests. Therefore I intentionnaly took a dive into generative audio models, focusing mainly on musical ouputs.
Keywords: SOTA Audio Generative models, Literature review, Deep Learning, Signal Processing
Text-Based Real Image Editing with Diffusion Models based on Google Imagic model
CVPR style research paperDescription: This project is based on this paper. The goal of this project was to understand the paper and try to reimplement parts of it. Using GCP, I re-trained and tested my model based on Imagic. I obtained quite satisfying results and tried also to explain any flaws in my work.
Keywords: Deep Learning, Diffusion Models, VAE, Artificial Vision, CVPR style research paper, Pytorch
Real-time movement-based music player using phones
Description: Inspired by the famous experience of Jacob Collier making the crowd sing 3-notes chord (youtube video), this project aimed to use smartphone movement data such as gyroscopical data to control a synthetizer chord.
Keywords: Max8/MSP, Deep Learning (RNN, LSTM), Python (torch), Data creation and augmentation, Smartphone live interactions.
Kaggle Bird Classification
Read the reportDescription: A Kaggle challenge where one needs to classify bird species given their picture. My approach was mainly to try to create a model from scratch instead of taking already trained models. This helped me to better understand the architectures of deep learning vision models.
Keywords: Deep Learning (CNN), Computer Vision, Python (torch), Data augmentation, Optimization.
Combining Radiance Transfer Method and Feedback Delay Networks for Late Reverberation Synthesis
Read the base paperDescription: A work on this paper. The main goal was to understand, and be able to re-explain what the authors did and why. I focused mainly on the theoratical computations. I did expose the paper strengths, weaknesses and openings. I also explored solutions involving deep learning techniques to improve the model.
Keywords: Digital Signal processing, Reverberation, Mathematical Proofs, Conference-style paper presentation, Explored DL Solutions for model improvement
Statistical study of insurance company assets and liabilities
Report not available yetDescription: This project was done in collaboration with Milliman. It explored statistical tests used in finance to assess the martingale property of discounted asset processes. Our group analyzed and compared several tests, including their strengths, weaknesses, and performance, with a focus on optimizing their application to financial data.
Keywords: statistics (R), Mathematical formalism, Collaboration with Milliman