Multi-Task Learning of Height and Semantics from Aerial Images

Apprentissage multi-tâche de l'élévation et de la sémantique à partir d'images aériennes

Marcela Carvalho (1) , Bertrand Le Saux (1) , Pauline Trouvé-Peloux (1) , Frédéric Champagnat (1) , Andrés Almansa (2)

1 DTIS, ONERA, Université Paris Saclay (COmUE) [Palaiseau]
2 MAP5 - UMR 8145 - Mathématiques Appliquées Paris 5

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 .

Keywords: AERIAL IMAGERY; DEEP LEARNING; MULTITASK LEARNING; NEURAL NETWORKS; SEMANTIC SEGMENTATION; SINGLE VIEW DEPTH ESTIMATION

Keywords: TELEDETECTION; APPRENTISSAGE AUTOMATIQUE; DEEP LEARNING; RESEAU NEURONES; CARTOGRAPHIE URBAINE; CLASSIFICATION; IMAGERIE HAUTE RESOLUTION; LIDAR AERIEN

Domains: Engineering Sciences [physics]; Physics [physics]; Mathematics [math]; Computer Science [cs]

Marcela Carvalho, Bertrand Le Saux, Pauline Trouvé-Peloux, Frédéric Champagnat, Andrés Almansa. Multi-Task Learning of Height and Semantics from Aerial Images. IEEE Geoscience and Remote Sensing Letters, 2019, ⟨10.1109/LGRS.2019.2947783⟩. ⟨hal-02386074v2⟩