Applied AI: on-board models, research datasets, and developer tools.
Vox builds AI-powered products and tools. Here on Hugging Face we publish work in two areas: models that run on board unmanned aircraft for disaster reconnaissance, and datasets for patent research and agent evaluation, including resources developed for Parallax. Explore our software and agent frameworks on GitHub.
Models published under this organization are developed in Japan by Vox, which operates from Tokyo.
Our on-board work targets the first hours after a flood or an earthquake, when a drone has to decide what to look at next rather than fly a route fixed before take-off. The models read the current view and emit a viewpoint sequence, which a separate safety layer checks against a pre-approved volume before anything reaches the flight controller. They are fine-tunes of open base models, trained on public UAV footage published by the Geospatial Information Authority of Japan under the Japanese Public Data License.
Model cards carry the measured results, the retracted ones, and the limits, including what has not been tested.
Read each model and dataset card for its scope, provenance, measurements, limitations, citation, and license terms. For questions, feedback, or data-quality reports, open a discussion in the relevant repository.