@incollection{fdi:010085560,
  title = {{A}ssessment of {CNN}-based methods for poverty estimation from satellite images},
  author = {{J}arry, {R}. and {C}haumont, {M}. and {B}erti-{E}quille, {L}aure and {S}ubsol, {G}},
  editor = {},
  language = {{ENG}},
  abstract = {{O}ne of the major issues in predicting poverty with satellite images is the lack of fine-grained and reliable poverty indicators. {T}o address this problem, various methodologies were proposed recently. {M}ost recent approaches use a proxy (e.g., nighttime light), as an additional information, to mitigate the problem of sparse data. {T}hey consist in building and training a {CNN} with a large set of images, which is then used as a feature extractor. {U}ltimately, pairs of extracted feature vectors and poverty labels are used to learn a regression model to predict the poverty indicators.{F}irst, we propose a rigorous comparative study of such approaches based on a unified framework and a common set of images. {W}e observed that the geographic displacement on the spatial coordinates of poverty observations degrades the prediction performances of all the methods. {T}herefore, we present a new methodology combining grid-cell selection and ensembling that improves the poverty prediction to handle coordinate displacement.},
  keywords = {},
  booktitle = {{P}attern {R}ecognition : proceedings, part {VII}},
  numero = {12667},
  pages = {550--565},
  address = {{C}ham},
  publisher = {{S}pringer},
  series = {{L}ecture {N}otes in {C}omputer {S}cience},
  year = {2021},
  DOI = {10.1007/978-3-030-68787-8_40},
  ISBN = {978-3-030-68786-1},
  URL = {https://www.documentation.ird.fr/hor/fdi:010085560},
}