Computer Vision News - July 2022

54 Congrats, Doctor! The unlimited access to earth observation data has resulted in the development of powerful algorithms able to survey the planet. Still, problems like misregistration, intraclass variation as well as complex spatial andspectral distributionsof ground objects continue to hinder the formulation of robust and generic algorithms. To this end, our research focused on creating an effective multi-task change detection algorithm based on sequential data (in our case time-series images) that takes full advantage of the temporal relationship between them. At the same time, a multi- stepdeformable image registrationmethod was developed, able to handle not only the global affine displacements but also the more complex local pixelwise deformations caused by the different viewpoints of the sensors. All our methods were examined proposing novel data-driven deep learning methods on high and very high resolution satellite images . Giving some more details on the change detection topic, the proposed algorithm [Figure1] is a deep multi-task learning framework able to couple semantic segmentation and change detection using fully convolutional Long Short-Term Memory (LSTM) networks . In particular, we present a UNet-like architecture which models the temporal relationship of spatial feature representations using integrated fully convolutional LSTM blocks on top of every encoding level. In this way, the network is able to capture the temporal relationship of spatial feature vectors in all encoding levels without the need to downsample or flatten them, forming an end-to-endtrainableframework.Moreover, we further enrich this architecture with an additional decoding branch that performs Maria Papadomanolaki recently completed her PhD at the Remote Sensing Lab at the National Technical University of Athens, collaborating also with the Mathématiques et Informatique pour la Complexité et les Syst ѐ mes (MICS) Lab in CentraleSupélec, Université Paris-Saclay. Her research focused mainly on the topics of image registration and urban change detection using satellite images with deep learning methods. Maria continues her research as a Postdoctoral Scholar in CentraleSupélec at the MICS Laboratory working on the medical imaging field following the methodological advancements on computational pathology. Congrats, Doctor Maria!