/ / / / / CV

Hello! I am Antoine, an incoming Machine Learning Engineer at . I recently graduated with a master's degree in mathematics and artificial intelligence from ENS Paris-Saclay. I also completed a research internship at Meta Meta Logo . Prior to that, I took a gap year during which I had the opportunity to be a visiting researcher at New York University, specifically at NYU Video Lab under the supervision of Yao Wang and Adeen Flinker. Before joining ENS Paris-Saclay, I completed the first year of my master's degree at Paris-Saclay University in mathematics and artificial intelligence. I also received a bachelor's degree from Gustave Eiffel University in mathematics and computer science.

My research interests are signal processing, computer vision and their applications to the medical field, when possible. I am also interested in pre-training and multimodal models.

Projects

Wasserstein Barycenter and its Application to Texture Mixing December 2023

In this study, we explore the work conducted by Julien Rabin et al. to perform texture mixing using Wasserstein Barycenter. We try to reproduce the results and propose new techniques to enhance the texture synthesis. [Paper] [Code]

Denoising score matching for diffusion models December 2023

Diffusion models have shown impressive results on many tasks such as image generation, inpainting or denoising. In this paper, we investigate how the score of a probability distribution \(p(x)\), characterized by \(\nabla_x \log(p(x))\), can be used for these tasks. [Paper]

Diffusion Det January 2023

DiffusionDet is an architecture made for object detection based on diffusion models developed by Soufa Chen. In this project, we tried to improve the object tracking system by using a prior distribution on the boxes from the previous frame. We then evaluated the performances on the MOT16 dataset. We also tried to implement SIFT for object tracking. [Paper]

Research

Critical regions prediction March 2023 - Present

At New York University, we are developing a model able to detect critical regions of the brain thanks to electrocorticography (ECoG). We define critical regions as the parts of the brain that are stimulated and required when we perform tasks such as speaking. The gold standard for identifying these regions has remained unchanged for many years and is known as electrical stimulation mapping (ESM). However, this technique of brain stimulation carries the risk of discharge and seizures in both diseased and healthy tissue. In addition, this is often not well tolerated by pediatrics. For these reasons, and many more we did not evoke, there is a crucial need to improve this method, which could be beneficial for pre-surgery planning to ensure critical regions are not inadvertently resected. This work is jointly supervised by Yao Wang and Adeen Flinker.

Classifying the state of a patient in intensive care unit

Classifying the state of sedation of patients in ICU November 2022 - January 2023

This study aimed to classify the state of sedation of patients during their hospitalizations in an intensive care unit. The doctor judged between a "sedated" or "weakly sedated" state for each patient and proposed medication according to this state. By studying the electroencephalographic activity (EEG) of the patients, we have tried to classify these states. This work was supervised by Laurent Oudre and Clément Dubost.

Land's kernel April 2022 - May 2022

Kernels play an important role in image processing and computer vision. They are ubiquitous in our daily lives, and have many applications such as feature extraction or photomontage. In this work, we will focus on Land’s kernel and its interesting properties. This work was supervised by Enric Meinhardt-Llopis. [Paper] [Code]

Antoine Ratouchniak

Projects

Research

Miscellaneous