I am a PhD Candidate in Image Processing and Computer Vision at ENS Paris-Saclay. I work on change detection on satellite imagery and video. I am interested in problems with limited annotations, i.e. unsupervised, few-shot, weakly and self-supervised learning methods.

News

• April 2024: Our paper Exploring Robust Features for Few-Shot Object Detection in Satellite Imagery has been accepted at the EarthVision Workshop at CVPR24.
• March 2024: Our paper Portraying the Need for Temporal Data in Flood Detection via Sentinel-1 has been accepted at IGARSS24.
• March 2023: I am visiting Guillermo Sapiro’s group at Duke University with a team led by Jean-Michel Morel.
• November 2022: I am starting my PhD at Centre Borelli, ENS Paris-Saclay, advised by Rafael Grompone and Thibaud Ehret.
• February 2022: I am joining the Centre Borelli as a research intern, advised by Jean-Michel Morel and Gabriele Facciolo.

Publications

	Structure Tensor Representation for Robust Oriented Object Detection Xavier Bou, Gabriele Facciolo, Rafael Grompone Von Gioi, Jean-Michel Morel, Thibaud Ehret Preprint, 2024 paper We propose to represent orientation as a structure tensor in Oriented Object Detection, bridging the gap between Gaussian-based and angle-coder solutions.
	Exploring Robust Features for Few-Shot Object Detection in Satellite Imagery Xavier Bou, Gabriele Facciolo, Rafael Grompone Von Gioi, Jean-Michel Morel, Thibaud Ehret CVPR Workshops, 2024 paper / code / poster We explore recent ideas on Open Vocabulary Detection to detect any object in remote sensing images with only a handful of examples.
	Portraying the Need for Temporal Data in Flood Detection via Sentinel-1 Xavier Bou, Thibaud Ehret, Rafael Grompone von Gioi, Jérémy Anger IGARSS, 2024 paper / code / poster We address current limitations in flood detection and illustrate the importance of temporal information to solve the flood detection problem, extending the MMFlood dataset to multi-date.
	Statistical modeling of deep features reduces false alarms in video change detection Xavier Bou, Aitor Artola, Thibaud Ehret, Gabriele Facciolo, Jean-Michel Morel, Rafael Grompone von Gioi Preprint, 2024 paper Introduces a weakly supervised a-contrario validation process, based on high dimensional statistical modeling of deep features, to reduce the number of false alarms of any change detection algorithm.
	Reviewing ViBe, a Popular Background Subtraction Algorithm for Real-Time Applications Xavier Bou, Thibaud Ehret, Gabriele Facciolo, Jean-Michel Morel, Rafael Grompone von Gioi IPOL, 2022 paper We review the classical video background subtraction algorithm ViBe, which is still widely due to its simplicity and low computational load. An interactive demo is provided to quickly run and visualize its results via an easy-to-use interface.
	A Study of RobustNet, a Domain Generalization Method for Semantic Segmentation Xavier Bou IPOL, 2022 paper A review of RobustNet, a Domain Generalization method for Urban-Scene Semantic Segmentation. Instead of exposing the network to a wide range of domains, RobustNet tries to separate domain-variant from domain-invariant features via a whitening transformation. A demo is provided to easily test RobustNet and see the results on your own data.