Welcome to the page presenting my projects. Discover a selection of technological projects I developed, their objectives, the challenges tackled, my contributions, and test the final results.

Image 1 Image 2 Image 3

[AI] Mini Diffusion Model - Demo available below!

Diffusion models are a class of probabilistic generative models that learn to generate data by reversing a process of progressive degradation (diffusion) from noise to real data. They are particularly effective for generating realistic images, videos and even textual data by simulating this process in a controlled, iterative way. They are used in particular by software such as Dall-E or Stable Diffusion.

This project is a first experiment to implement a 'mini' diffusion model (these models generally have too many parameters to be developed and trained from scratch by an individual) that generates faces according to physical attribute conditions.


Objectives & Contributions:

  • Rewriting and local training of a Diffusion Model
  • Hosting of a model and creation of an online demo interface
  • Understanding the theoretical framework of diffusion models

Potential next steps:

  • Understanding and experimenting with fine-tuning techniques on foundation models (e.g., LoRA)
  • Implementation and testing of Latent Diffusion
  • Understanding and experimenting with text-to-image generation techniques

Technologies & Skills: Python, PyTorch, Numpy, Pandas, Git, Gradio, Hugging Face Spaces, CUDA, Diffusion Model, Supervised learning, Unet Architecture.

Useful links :

Image 1

[Web] My personal Website - Work In Progress

Quite simply the site you're currently browsing. I undertook this project mainly to have a medium on which to display interactive demonstrations of my work, projects and experiences, as well as to be able to give my profile a personal touch. As web development is not my speciality, I limited myself to the front-end of the site.


Objectives & contributions:

  • Develop an interactive front-end interface (zoom effects, carousel, etc.)
  • Host AI models and integrate an online demonstration tool

Technologies & Skills: JavaScript, CSS, HTML, Prompt Engineering, Git.

Useful links:

Image 1 Image 2

[AI] Video Frame Interpolation

Video frame interpolation consists of determining a new image between 2 adjacent images in a video. Its many applications include increasing the number of frames (when restoring old videos or smoothing out stop-motion videos, for example), video slow-motion and smoothing out video streams (television, streaming). I did this project as the final output of my 'Machine Learning & differentiable programming' course at Mines Nancy (15-20 hours), which was my first Deep Learning project and my first practical experience of PyTorch.


Objectives & Contributions :

  • Study the state of the art in video frame interpolation
  • Writing, training and evaluating a first Deep Learning model from scratch
  • Use parallel computing tools
  • Practical experience in handling video data

Technologies & Skills: Python, PyTorch, Numpy, Git, CUDA, Video Frame Interpolation, Unet Architecture.

Useful links: