Vous êtes ici

Semi-supervised semantic segmentation for large-scale automated cartography

Equipe et encadrants
Département / Equipe: 
Site Web Equipe: 
http://www.irisa.fr/obelix
Directeur de thèse
Sébastien Lefèvre
Co-directeur(s), co-encadrant(s)
Bertrand Le Saux
Alexandre Boulch
Contact(s)
NomAdresse e-mailTéléphone
Sébastien Lefèvre
sebastien.lefevre@irisa.fr
0297017266
Sujet de thèse
Descriptif

The evergrowing satellite imagery data in the last two decades have allowed new developments in the fileds of ecology, urban planning or natural disaster response. Those data are also more easily available, even openly as in the Copernicus program of the European Space Agency (Sentinel satellites). However, data exploitation requires human interprets, for example to identify tree species to study deforestation in a local ecosystem or find new buildings to measure growth of urban areas.

Thanks to the new deep learning methods developped for processing multimedia images in recent years, it now becomes possible to automate most of these processings for Earth-observation data. Indeed, many state-of-the-art algorithms for object detection and image segmentation or classification [Audebert 2016, Rey 2017] have been successfully transfered for aerial and satellite images. It allowes to produce quickly and without human intervention precise semantic mappings, in both urban and rural contexts.

This thesis project aims at large-scale, automated cartography, which raise several problems.

Problem 1: How to make semantic segmentation on a real large scale ?

Problem 2: How to deal with sparse, heterogeneous data?

Problem 3: How to predict maps with varying semantic levels?

Problem 4: How to estimate which prediction is possible depending on the available data?

This thesis will be carried out as part of collaboration between ONERA/DTIS (Bertrand Le Saux and Alexandre Boulch) and the OBELIX team of IRISA at University Bretagne Sud (Sébastien Lefèvre, PhD supervisor). It will take place mainly in the ONERA centre in Palaiseau (near Paris).

See https://www-obelix.irisa.fr/job-offers/

Bibliographie

[Audebert 2016] Semantic Segmentation of Earth Observation Data using Multimodal and Multi-scale Deep Networks; Audebert, N., Le Saux, B., Lefèvre, S.; Asian Conference on Computer Vision (ACCV), 2016.

[Audebert 2017a] Segment-before-Detect: Vehicle Detection and Classification through Semantic Segmentation of Aerial Images; Audebert, N., Le Saux, B., Lefèvre, S.; Remote Sensing, 2017.

[Audebert 2017b] Joint Learning from Earth Observation and OpenStreetMap Data to Get Faster Better Semantic Maps; Audebert, N., Le Saux, B., Lefèvre, S.; EarthVision - CVPR Workshop, 2017.

[Durand 2017]  WILDCAT: Weakly Supervised Learning of Deep ConvNets for Image Classification, Pointwise Localization and Segmentation; Durand, T., Mordan, T.,  Thome, N.,  Cord, M.; IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017

[Khoreva 2017]  Simple Does It: Weakly Supervised Instance and Semantic Segmentation; Khoreva, A., Benenson, R., Hosang, J., Hein, M., Schiele, B.; IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017

[Krizhevsky 2012] Imagenet classification with deep convolutional neural networks; Krizhevsky, A., Sutskever, I., Hinton, G.E.; Advances in neural information processing system (NIPS), 2012

[Lefèvre 2017]  Toward Seamless Multiview Scene Analysis From Satellite to Street Level; Lefèvre, S., Tuia, D., Wegner, J.D., Produit, T., Nassaar, A.S.; Proceedings of the IEEE, 2017

[Maggiori 2017] Can Semantic Labeling Methods Generalize to Any City? The Inria Aerial Image Labeling Benchmark; Maggiori, E., Tarabalka, Y., Charpiat, G., Alliez, P.; IEEE International Geoscience and Remote Sensing Symposium (IGARSS), 2017

[Mnih 2012] Learning to Label Aerial Images from Noisy Data; Mnih, V., Hinton, G.; International Conference on Machine Learning (ICML), 2012

[Rasmus 2017] Semi-Supervised Learning with Ladder Networks; Rasmus, A., Valpola, H., Honkala, M., Berglund, M., Raiko T.; International Conference on Neural Information Processing Systems (NIPS), 2015

[Redmon 2017] YOLO9000: Better, Faster, Stronger; Redmon, J. and Fahradi, A.; IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017

[Rey 2017]  Detecting animals in African Savanna with UAVs and the crowds; Rey, N., Volpi, M., Joost, S. and Tuia, D.; Remote Sensing of Environment, 2017

[Sumbul 2017] Fine-Grained Object Recognition and Zero-Shot Learning in Remote Sensing Imagery; Sumbul, G., Cinbis, R.G., Aksoy, S.; IEEE Transactions on Geosciences and Remote Sensing, 2017

Début des travaux: 
As soon as possible
Mots clés: 
deep learning, semi-supervised learning, remote sensing
Lieu: 
ONERA, Palaiseau