Skip to Main content Skip to Navigation
Conference papers

Data annotation with active learning: application to environmental surveys

Chloé Friguet 1, 2 Romain Dambreville 1 Ewa Kijak 3 Mathieu Laroze 1 Sébastien Lefèvre 1, 2
1 OBELIX - Environment observation with complex imagery
UBS - Université de Bretagne Sud, IRISA-D5 - SIGNAUX ET IMAGES NUMÉRIQUES, ROBOTIQUE
3 LinkMedia - Creating and exploiting explicit links between multimedia fragments
Inria Rennes – Bretagne Atlantique , IRISA-D6 - MEDIA ET INTERACTIONS
Abstract : An active learning framework is introduced to deal with reducing the annotation cost for aerial images in environmental surveys. The selection of the queried instances at each step of the active process is here constrained by requiring that they belong to a group, an image (or a part of it) in our case. A score to rank the images and identify the one that should be annotated at each iteration is defined, based on both classifier uncertainty and performances. The performances of several strategies regarding the interaction gain are discussed based on an experiment on real image data collected for an environmental survey.
Complete list of metadata

https://hal.archives-ouvertes.fr/hal-03445740
Contributor : Chloé Friguet Connect in order to contact the contributor
Submitted on : Wednesday, November 24, 2021 - 10:22:14 AM
Last modification on : Friday, November 26, 2021 - 3:34:47 AM

File

JDS_20_VF2.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03445740, version 1

Citation

Chloé Friguet, Romain Dambreville, Ewa Kijak, Mathieu Laroze, Sébastien Lefèvre. Data annotation with active learning: application to environmental surveys. Journées de Statistique (SFdS), May 2020, Nice, France. ⟨hal-03445740⟩

Share

Metrics

Record views

21

Files downloads

4