Welcome to SharingHub¶
Welcome to our AI-focused web portal, SharingHub, designed to help you discover, navigate, and analyze your AI-related Git projects hosted on GitLab. Here you can download code, models and datasets with our interface focused on navigation and accessibility!
We take inspiration from Hugging Face, the most used AI sharing website in the world, but our goal is turned towards the geospatial component, with GIS capabilities, with datasets and AI models referenced in space and time. This portal is intended to improve exchanges and collaboration between scientists by integrating voting mechanisms, code reviews, and the complete management of the shared assets life cycle.
Features¶
Warning
Because SharingHub itself is configurable, and rely on your GitLab instance, some features may not be available for your instance.
GitLab integration¶
Very light, SharingHub is deployed on top of a GitLab instance, and dynamically extracts metadata from the shared projects to serve them through its web interface, as well as a normalized API.
Standardized API¶
Our API implements the STAC specification, and more precisely the STAC API. This specification is strongly integrated to OGC standards, and we are compliant with the OGC API Features.
Data storage¶
We can deliver large datasets by offering a DVC remote (S3 store), and support of Git LFS for compatible GitLab instances.
Machine learning tracking¶
With its integration with MLflow, the SharingHub can be used for remote tracking of ML training. The tracking is directly linked to the repository containing the model, to keep the training runs close to the source code.