@article{fdi:010091575,
  title = {{D}eep{W}ealth : a generalizable open-source deep learning framework using satellite images for well-being estimation},
  author = {{B}en {A}bbes, {A}. and {M}achicao, {J}. and {C}orr{\^e}a, {P}.{L}.{P}. and {S}pecht, {A}. and {D}evillers, {R}odolphe and {O}metto, {J}.{P}. and {K}ondo, {Y}. and {M}ouillot, {D}.},
  editor = {},
  language = {{ENG}},
  abstract = {{M}easuring socioeconomic indices at the scale of regions or countries is required in various contexts, in particular to inform public policies. {T}he use of {D}eep {L}earning ({DL}) and {E}arth {O}bservation ({EO}) data is becoming increasingly common to estimate specific variables like societal wealth. {T}his paper presents an end-to-end framework '{D}eep{W}ealth' that calculates such a wealth index using open-source {EO} data and {DL}. {W}e use a multidisciplinary approach incorporating satellite imagery, socio-economic data, and {DL} models. {W}e demonstrate the effectiveness and generalizability of {D}eep{W}ealth by training it on 24 {A}frican countries and deploying it in {M}adagascar, {B}razil and {J}apan. {O}ur results show that {D}eep{W}ealth



provides accurate and stable wealth index estimates with an of 0.69. {I}t empowers computer-literate users skilled in {P}ython and {R} to estimate and visualize well-being-related data. {T}his open-source framework follows {FAIR} ({F}indable, {A}ccessible, {I}nteroperable, {R}eusable) principles, providing data, source code, metadata, and training checkpoints with its source code made available on {Z}enodo and {G}it{H}ub. {I}n this manner, we provide a {DL} framework that is reproducible and replicable.},
  keywords = {{ANGOLA} ; {BENIN} ; {BURKINA} {FASO} ; {CAMEROUN} ; {COTE} {D}'{IVOIRE} ; {ETHIOPIE} ; {GHANA} ; {GUINEE} ; {KENYA} ; {LESOTHO} ; {MALAWI} ; {MALI} ; {MOZAMBIQUE} ; {NIGERIA} ; {MADAGASCAR} ; {RWANDA} ; {SENEGAL} ; {SIERRA} {LEONE} ; {TANZANIE} ; {TOGO} ; {OUGANDA} ; {ZAMBIE} ; {ZIMBABWE} ; {REPUBLIQUE} {DEMOCRATIQUE} {DU} {CONGO}},
  booktitle = {},
  journal = {{S}oftware{X}},
  volume = {27},
  numero = {},
  pages = {101785 [8 ]},
  ISSN = {2352-7110},
  year = {2024},
  DOI = {10.1016/j.softx.2024.101785},
  URL = {https://www.documentation.ird.fr/hor/fdi:010091575},
}