2025-11-15 –, Way
With MapTCHA we explore if the validation of computationally generated predictions of map features could serve as an image based CAPTCHA to map new OSM objects. Following an earlier proof-of-concept at FOSDEM 2025, here we provide a development update.
CAPTCHAs are widely used by websites to stop automated registrations or scrapers. Often, they act as data-crowdsourcing at the same time, e.g. for image recognition as in Google's reCAPTCHA.
With MapTCHA, we are evaluating whether such an approach could help improve OpenStreetMap. Through efforts like mapathons and mapping initiatives led by the Humanitarian OpenStreetMap Team (HOT), volunteers of varying levels of experience create large amounts of objects for OSM. Similarly, there are efforts to use image recognition to create potential OSM additions. Both of these require extra, labour-intensive validations.
The idea of MapTCHA is to explore if we can present website visitors with both unvalidated and pre-validated sets of satellite images that include human/machine-drawn object annotations, currently buildings, asking them to mark the ones that are correctly outlined.
At FOSDEM 2025 we presented an early prototype to demonstrate our general idea. Here, we present some first data on the validity and accuracy of such crowd-sourced validations. We compare it to data that has already been validated through the use of MapSwipe, a similar crowdsourcing-based data validation tool developed by the MissingMaps project in collaboration with HeiGit.
Crowdsourcing, humanitarian mapping, CAPTCHA, open source, image recognition.
Affiliation:The MapTCHA team
A maps passionate currently employed as a Research Fellow at the Alan Turing Institute, in London, in the Urban Analytics team. As part of this role, Anna has carried out academic research to investigate the accuracy of a computer vision tool developed by HOT (Humanitarian OpenStreetMap Team) to map buildings from satellite imagery, and is currently working on this topic in collaboration with HeiGit (Heidelberg Institute for Geoinformation Technology, Germany).
Bastian is an interdisciplinary researcher working on the theory & practice of peer-production and citizen science, he has co-led and contributes to a wide range of FLOSS efforts covering many domains. You can learn more about him and his work at https://tzovar.as