Multitask Learning of Height and Semantics from Aerial Images

Marcela Carvalho; Bertrand Le Saux; Pauline Trouve-Peloux; Frederic Champagnat; Andres Almansa

doi:10.1109/LGRS.2019.2947783

Multitask Learning of Height and Semantics from Aerial Images

Marcela Carvalho
, Bertrand Le Saux
, Pauline Trouve-Peloux
, Frederic Champagnat
, Andres Almansa

Université Paris-Saclay
Laboratoire de Probabilités et Modèles Aléatoires

Research output: Contribution to journal › Article › peer-review

64 Citations (Scopus)

Abstract

Aerial or satellite imagery is a great source for land surface analysis, which might yield land-use maps or elevation models. In this letter, we present a neural network framework for learning semantics and local height together. We show how this joint multitask learning benefits to each task on the large data set of the 2018 Data Fusion Contest. Moreover, our framework also yields an uncertainty map that allows assessing the prediction of the model. Code is available at http://github.com/marcelampc/mtl_aerial_images

Original language	English
Article number	8891800
Pages (from-to)	1391-1395
Number of pages	5
Journal	IEEE Geoscience and Remote Sensing Letters
Volume	17
Issue number	8
DOIs	https://doi.org/10.1109/LGRS.2019.2947783
Publication status	Published - 1 Aug 2020
Externally published	Yes

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 15 Life on Land

Keywords

Aerial imagery
deep learning
multitask learning
neural networks
semantic segmentation
single view depth estimation

Access to Document

10.1109/LGRS.2019.2947783

Cite this

@article{3a3a0cda6ef841dabd517606ccac46d1,

title = "Multitask Learning of Height and Semantics from Aerial Images",

abstract = "Aerial or satellite imagery is a great source for land surface analysis, which might yield land-use maps or elevation models. In this letter, we present a neural network framework for learning semantics and local height together. We show how this joint multitask learning benefits to each task on the large data set of the 2018 Data Fusion Contest. Moreover, our framework also yields an uncertainty map that allows assessing the prediction of the model. Code is available at http://github.com/marcelampc/mtl\_aerial\_images",

keywords = "Aerial imagery, deep learning, multitask learning, neural networks, semantic segmentation, single view depth estimation",

author = "Marcela Carvalho and \{Le Saux\}, Bertrand and Pauline Trouve-Peloux and Frederic Champagnat and Andres Almansa",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.",

year = "2020",

month = aug,

day = "1",

doi = "10.1109/LGRS.2019.2947783",

language = "English",

volume = "17",

pages = "1391--1395",

journal = "IEEE Geoscience and Remote Sensing Letters",

issn = "1545-598X",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

number = "8",

}

TY - JOUR

T1 - Multitask Learning of Height and Semantics from Aerial Images

AU - Carvalho, Marcela

AU - Le Saux, Bertrand

AU - Trouve-Peloux, Pauline

AU - Champagnat, Frederic

AU - Almansa, Andres

PY - 2020/8/1

Y1 - 2020/8/1

N2 - Aerial or satellite imagery is a great source for land surface analysis, which might yield land-use maps or elevation models. In this letter, we present a neural network framework for learning semantics and local height together. We show how this joint multitask learning benefits to each task on the large data set of the 2018 Data Fusion Contest. Moreover, our framework also yields an uncertainty map that allows assessing the prediction of the model. Code is available at http://github.com/marcelampc/mtl_aerial_images

AB - Aerial or satellite imagery is a great source for land surface analysis, which might yield land-use maps or elevation models. In this letter, we present a neural network framework for learning semantics and local height together. We show how this joint multitask learning benefits to each task on the large data set of the 2018 Data Fusion Contest. Moreover, our framework also yields an uncertainty map that allows assessing the prediction of the model. Code is available at http://github.com/marcelampc/mtl_aerial_images

KW - Aerial imagery

KW - deep learning

KW - multitask learning

KW - neural networks

KW - semantic segmentation

KW - single view depth estimation

UR - https://www.scopus.com/pages/publications/85089179800

U2 - 10.1109/LGRS.2019.2947783

DO - 10.1109/LGRS.2019.2947783

M3 - Article

AN - SCOPUS:85089179800

SN - 1545-598X

VL - 17

SP - 1391

EP - 1395

JO - IEEE Geoscience and Remote Sensing Letters

JF - IEEE Geoscience and Remote Sensing Letters

IS - 8

M1 - 8891800

ER -

Multitask Learning of Height and Semantics from Aerial Images

Abstract

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this