1. Apprentissage multi-tâche de l'élévation et de la sémantique à partir d'images aériennes
- Author
-
Marcela Carvalho, Pauline Trouvé-Peloux, Frédéric Champagnat, Bertrand Le Saux, Andrés Almansa, DTIS, ONERA, Université Paris Saclay (COmUE) [Palaiseau], ONERA-Université Paris Saclay (COmUE), Mathématiques Appliquées Paris 5 (MAP5 - UMR 8145), Université Paris Descartes - Paris 5 (UPD5)-Institut National des Sciences Mathématiques et de leurs Interactions (INSMI)-Centre National de la Recherche Scientifique (CNRS), DTIS, ONERA, Université Paris Saclay [Palaiseau], ONERA-Université Paris-Saclay, and Université Paris Saclay (COmUE)-ONERA
- Subjects
FOS: Computer and information sciences ,Computer Science - Machine Learning ,RESEAU NEURONES ,Computer science ,Computer Vision and Pattern Recognition (cs.CV) ,0211 other engineering and technologies ,Computer Science - Computer Vision and Pattern Recognition ,IMAGERIE HAUTE RESOLUTION ,Multi-task learning ,02 engineering and technology ,NEURAL NETWORKS ,Semantics ,Machine learning ,computer.software_genre ,CLASSIFICATION ,Machine Learning (cs.LG) ,SEMANTIC SEGMENTATION ,[SPI]Engineering Sciences [physics] ,CARTOGRAPHIE URBAINE ,SINGLE VIEW DEPTH ESTIMATION ,Satellite imagery ,[INFO]Computer Science [cs] ,Electrical and Electronic Engineering ,[MATH]Mathematics [math] ,TELEDETECTION ,021101 geological & geomatics engineering ,AERIAL IMAGERY ,[PHYS]Physics [physics] ,APPRENTISSAGE AUTOMATIQUE ,Artificial neural network ,business.industry ,DEEP LEARNING ,Deep learning ,15. Life on land ,Geotechnical Engineering and Engineering Geology ,Sensor fusion ,Data set ,MULTITASK LEARNING ,Artificial intelligence ,LIDAR AERIEN ,business ,computer - Abstract
Aerial or satellite imagery is a great source for land surface analysis, which might yield land use maps or elevation models. In this investigation, we present a neural network framework for learning semantics and local height together. We show how this joint multi-task learning benefits to each task on the large dataset of the 2018 Data Fusion Contest. Moreover, our framework also yields an uncertainty map which allows assessing the prediction of the model. Code is available at https://github.com/marcelampc/mtl_aerial_images ., Comment: Published IEEE Geoscience and Remote Sensing Letters. Code https://github.com/marcelampc/mtl_aerial_images
- Published
- 2019