Bots and spam challenges for online platforms. Traditional captchas can help block bots, but often involves improve proprietary maps and software, while revealing user information to owners of owner-party captcha troviders. OpenStreetMap (OSM) has many items left map-mapped, but the quality of things AI has not enough participation. We introduced “Mapcha”, a captcha that fits uncertainty in interpreting computer visualization, users are asked to identify images containing images contained.
We separate the known positive cases, which prophecy and OSM contain an object, from unknown cases, which things are only in things. We also make negatives from places where things are not in OSM or prediction. We showed a mix of these images that the user does not say where. People are validated by confirming the known positives and negatives, and we determine the fact of unknown images through unknown images through unknown images through unknown images by accumulating images by compiled to voting users by voting. When voting indicates high confidence that something exists, we suggest location for OSM mapping.
Our prototype introduces buildings using aerial imagery with high resolution to imagine individual buildings and medium-sized items. Image recognition is given to the fair, an open source assisted mapping system developed by the humanitarian opensteretmap team (heat). It allows the training and good tuning of models that own the engine to claim the steps to spare treads.
Future plans include expanding many things and distinct to imagine; again with AI models; Join Mapcha with different login systems; and enhance user interface.
In this session we will discuss how we build this solution, how it can enhance the efforts to dismiss OSM to support AI data projects.
https://fosdem.org/2025/assets/style/logo-gear-7204a6874eb0128932db10ff4030910401ac06f4e907f8b4a40da24ba592b252.png
2025-02-13 22:44:00
