Team
MAMBO
Team manager: Weiss Pierre
Presentation
Our team mixes mathematics and computer science for:
– The improvement of microscopes through modeling, deep learning and computational imaging.
– The automatized analysis of 2D/3D biological images.
– The design and analysis of mathematical models in biology.
– Deriving fundamental mathematical results and certified algorithms. One of our main objectives lately has been to understand and probe the strengths and limits of “deep learning’’ and neural networks
– Developing open-source packages and plugins such as DeepInv, GlobalBioIm, Svetlana, Sketchpose, VSNR, …
Project 1
In this projet, we aim at recovering a space varying point spread function directly from 2D/3D microscope images. This serves multiple objectives :
– Characterize the optical system
– Improve the image resolution with computational techniques
– Design computational imaging systems (e.g. improve Random Illumination Microscopy) with the imaging platform
– Evaluate diffraction by the sample itself and recover biophysical properties such as refractive indices
Project 2
In this joint project with the team of Fabian Erdel, we aim at better understanding how condensates are formed within cells, with FRAP imaging techniques. After photobleaching, fluorescent molecules move and progressively recover a stationary state. By analyzing image sequences during the recovery, we can evaluate biophysical parameters such as the permeability of the boundary, diffusion coefficients inside and outside the condensate, probability to bind with other elements… This requires fine physics modeling, numerical algorithms and solution certification.
Project 3
In this project, we aim at developing new deep learning models to ease the segmentation and classification of objects with biological images. We are particularly interested in interactive solutions, where the user progressively teaches a deep learning model to accomplish a specific task.
Project 4
We perform basic research in mathematics to better understand the properties of complex objects encountered in mathematics, such as neural networks, inverse problems. This process nurtures all the rest of our research.
Project 5
Mathematical modeling is a good way to better understand living systems, clarify hypotheses, but also evaluate quantitative parameters from experiments. However, most of the time, the parameters can only be observed indirectly and incompletely. The goal of this project is to develop tools that automatically detect what parameters can be identified from an experiment coupled with a mathematical model and what parameters cannot. In a sense, we want to identify the non identifiable and guide biologists for new experiments.
– Sketchpose: learning to segment cells with partial annotations. C. Cazorla, N. Munier, R. Morin, P. Weiss. Preprint, (2024)
– Surpassing Light Inhomogeneities in Structured-Illumination Microscopy with FlexSIM. E. Soubies, A. Nogueron, F. Pelletier, T. Mangeat, C. Leterrier, M. Unser, and D. Sage. Journal of Microscopy, (2024)
– DEEP-BLUR: Blind Identification and Deblurring with Convolutional Neural Networks V. Debarnot, P. Weiss. Biological Imaging, Cambridge University Press, (2024)
– SVETLANA: A Supervised Segmentation Classifier for NAPARI. C. Cazorla, R. Morin, P. Weiss. Scientific Reports, (2024).
– Training Adaptive Reconstruction Networks for Blind Inverse Problems. A. Gossard, P. Weiss. SIAM Imaging Science, (2024).
– DeepVibes: Correcting Micro-Vibrations in Satellite Imaging with Pushbroom Cameras M.H. Nguyen, F. De Vieilleville, P. Weiss. IEEE Transactions on Geoscience and Remote Sensing, (2024).
– The MLE is a reliable source: sharp performance guarantees for localization problems. N. Munier, E. Soubies & P. Weiss, Inverse Problems, (2023).
– Bayesian Optimization of Sampling Densities in MRI. A. Gossard, F. de Gournay & P. Weiss, MELBA: Machine Learning for Biomedical Imaging, (2023).
– Sampling rates for l1 synthesis. M.März, C. Boyer, J. Kahn, P. Weiss. Foundations of Computational Mathematics, (2022).
– Learning low-dimensional models of microscopes. V. Debarnot, P. Escande, T. Mangeat, P. Weiss. IEEE Transactions on Computational Imaging, (2021).
Affiliation
