#co_loginIFrame { width:100%; height: 20em; } @media (max-width: 575px) { #co_loginIFrame { height: 46em; } } @media (min-width: 576px) { #co_loginIFrame { height: 48em; } } @media (min-width: 768px) { #co_loginIFrame { height: 34em; } } @media (min-width: 992px) { #co_loginIFrame { height: 26em; } } window.addEventListener("message", (event) => { let tURL = "https://meetingorganizer.copernicus.org/EGU22/session/43183"; if(event.origin.search(".copernicus.org")) { let data = event.data; } return; }, false);

SC5.3

Image analytics for crowdsourced photographs: no-code supervised and unsupervised classification solutions

Convener: Oleksandr KarasovECSECS | Co-convener: Evelyn Uuemaa

Geotagged photographs provide unique insights into the plurality of geographic tasks - from land use/land cover classifications to biodiversity, hazardous events, and people's well-being monitoring. Millions of passively and actively crowdsourced geotagged photographs are available online in open access via social media platforms (Twitter, Flickr, VK.com) and citizen science initiatives. However, handling big photo datasets requires advanced data science skills. This short course presents the democratised digital tools for image analytics and classification: TensorFlow deep learning classification model, implemented in Microsoft's Lobe software, and image analytics tools in Orange data mining software. Participants will learn the basics of image analytics, the difference between supervised and unsupervised classification, quality assessment and optimisation techniques for deep learning image classification models.