Note
This repository is still in the testing phase. Some features may be broken or unstable.
CRAB is a FOSS stack for ML dataset management and annotation. It includes preconfigured profiles for importing IFCB and LISST-Holo data. If you have another data format you would like added, please open an issue on GitHub or send me an email. CRAB manages ingesting data, providing all files in a uniform TIFF format on an S3 compatible object store. Metadata is searchable using queries to the integrated CouchDB server, allowing you to easily access data stored on the platform.
CRAB requires Docker Compose to run, but otherwise comes with everything you need for a minimal install out of the box.
For a local install, run flask/build.sh first.
Then run generate-env.sh to auto-generate a .env file.
If you're using externally managed S3 compatible storage, edit the .env file now and skip running init-db.sh. You'll need to configure databases manually.
Run init-db.sh to prepare your databases automatically.
Finally, run launch.sh to start your instance.
CRAB should now be avaliable on http://localhost:8080
For futher notes on installation, using the software, and connecting to the CRAB API, head to the docs folder. These markdown documents are also avaliable on an installed instance under the "Help" tab.
CRAB is free software under the AGPL-3.0 license. The CRAB Logo (© 2024, Alex Baldwin) is licensed under CC BY-SA 4.0.