Reproducible research
Research | Links:
I am heavily invested in the IPOL (Image Processing On Line) journal and demo system for open science and reproducible research, where scientists can create online demos of existing methods and publish companion technical articles describing the implementation. In particular, I cofounded and have been organizing the IPOL MLBriefs workshop and hackhathon since its creation in April 2022.
In the fourth edition, which was held at the ENS Paris-Saclay in May 2024, I had the honour to invite three plenary speakers to share their views on the modern challenges of machine learning and computer vision:
- Christian Sandor (LISN, ARAI team leader)
- Tomasz Bednarz (Director of Strategic Researcher Engagement at NVIDIA Corporation)
- Gaël Varoquaux (Inria Soda leader, cofounder of scikit-learn)
In many fields of applied mathematics and computer science (computer vision, machine learning, image/signal processing, …), new methods and applications are appearing at a vertiginous pace, to the point it is hard to keep up with all new methods. Yet, being able to understand and try these methods oneself is crucial for research.
To bridge this gap and promote reproducible research, we launched the IPOL MLBriefs workshop and hackathon in April 2022, to help people create and publish an online demo of a method on the IPOL journal and demo system, with a companion technical article describing the method and experimentally highlighting its use cases and limitations.
The MLBriefs workshop has already been held four times, the most recent edition took place in May, 2024.
Besides the workshop, I also participate in the IPOL editorial committees and contribute to the development of the IPOL demo system. In particular, I designed a local demo runner in Rust for IPOL.
Currently, I am writing a full tutorial on IPOL demo creation. Feel free to check back! In the meantime, please see the MLBriefs website for the recorded demo creation tutorials I gave at the workshops.
