3rd year PHD student at Télécom Paris, Institut Polytechnique de Paris. My research interests are image restoration, deep learning, and computer vision.
Deep neural networks have recently surpassed other image restoration methods which rely on hand-crafted priors. However, such networks usually require large databases and need to be retrained for each new modality. In this paper, we show that we can reach near-optimal performances by training them on a synthetic dataset made of realizations of a dead leaves model, both for image denoising and super-resolution. The simplicity of this model makes it possible to create large databases with only a few parameters. We also show that training a network with a mix of natural and synthetic images does not affect results on natural images while improving the results on dead leaves images, which are classically used for evaluating the preservation of textures. We thoroughly describe the image model and its implementation, before giving experimental results.
Deep neural networks models are generally designed and trained for a specific type and quality of data. In this work, we address this problem in the context of nested learning. For many applications, both the input data, at training and testing, and the prediction can be conceived at multiple nested quality/resolutions. We show that by leveraging this multi-scale information, the problem of poor generalization and prediction overconfidence, as well as the exploitation of multiple training data quality, can be efficiently addressed. We evaluate the proposed ideas in six public datasets: MNIST, Fashion-MNIST, CIFAR10, CIFAR100, Plantvillage, and DBPEDIA. We observe that coarsely annotated data can help to solve fine predictions and reduce overconfidence significantly. We also show that hierarchical learning produces models intrinsically more robust to adversarial attacks and data perturbations.
Standard deep neural networks (DNNs) are commonly trained in an end-to-end fashion for specific tasks such as object recognition, face identification, or character recognition, among many examples. This specificity often leads to overconfident models that generalize poorly to samples that are not from the original training distribution. Moreover, such standard DNNs do not allow to leverage information from heterogeneously annotated training data, where for example, labels may be provided with different levels of granularity. To address these challenges, we introduce the concept of nested learning: how to obtain a hierarchical representation of the input such that a coarse label can be extracted first, and sequentially refine this representation, if the sample permits, to obtain successively refined predictions, all of them with the corresponding confidence.
6 months Master internship under the supervision of Pr. Guillermo Sapiro and Pr. Matias Di Martino. During this internship, we jointly worked on a theoretical deep learning problem. The goal was to propose a novel way of tackling classification problems, by enforcing nested predictions. Most classification tasks can indeed be split in a nested taxonomy, which we exploit by providing multi-granular outputs, with the appropriate confidence. This projected resulted in a paper which shoud be available soon.
Courses followed : Image Denoising, Deep Learning, Mathematical foundations of Deep Learning, Mathematics of Imaging, Deep Leartning for Image Restoration, Online Learning, Reinforcement Learning, Object Recognition, Convex Optimization, Medical Imaging
Engineering degree in Telecommunications. The main topics studied are signal processing, computer science, and maths.
Apart from academical works, I really enjoy photography, especially when I travel. I mostly do street photography in Paris, and sometimes in Berlin, New York, Rome, etc. I shall release a gallery of those pictures soon.
Other than that, I always enjoy to take a dip in the over-crowded Parisian swimming pools, since I "retired" from my almost non-existing swimming career.