sharkLinq bundles the Shark Matcher inference service, the Shark Matcher web
application, and setup instructions for the sample dataset.
sharkLinq/
├── LICENSE # MIT License for the repository
├── sm-data/ # sample or working data (downloaded from GitHub Releases)
├── shark-match-model/ # ML inference service (training, embeddings, retrieval API)
└── shark-matcher-app/ # web application and SQLite database tooling
Both shark-match-model and shark-matcher-app automatically look for
sm-data/ as a sibling directory. To store sm-data/ elsewhere, set
SHARK_DATA_ROOT to its absolute path in every terminal running either service.
git clone REPOSITORY_LINK
cd sharkLinqNote
The sample download already includes formatted images, database CSVs,
precomputed embeddings (pre2019-sample-data), and a trained model checkpoint
(omega_miew_arcface_random). You do not need to clean data, train a model,
or generate embeddings to run the app demo with the sample data.
-
Download
sm-data.zipfrom the latest release onREPOSITORY_LINK/releasesand extract it in thesharkLinq/root directory:curl -L -o sm-data.zip REPOSITORY_LINK/releases/latest/download/sm-data.zip unzip sm-data.zip rm sm-data.zip
Or with the GitHub CLI:
gh release download --repo REPOSITORY_LINK --pattern "sm-data.zip" unzip sm-data.zip rm sm-data.zip -
Confirm
sm-data/sits besideshark-match-model/andshark-matcher-app/:ls sm-data
Expected directories include
annotations,database-format,embeddings,images, andtrained-models.
Follow the Getting Started section in shark-matcher-app/README.md to
create the Python 3.12 virtual environments, install dependencies, populate the
database, and start both services.
- For inference service details—or optional data cleaning, model training, and
embedding generation on your own data—see
shark-match-model/README.md. - For the interactive matching demo (
--demo), seeshark-matcher-app/demo/README.md.
sharkLinq (shark-match-model and shark-matcher-app) is released under the
MIT License by the LINQS Lab at UC Santa Cruz; see the LICENSE file.
The images and annotations in the demo bundle come from a public dataset licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). Credit the authors whenever you reuse them:
- Paul E. Kanive, Salvador J. Jorgensen, Jay J. Rotella, Barbara A. Block, Taylor K. Chapple, Scot D. Anderson, & Timothy White. (2020). Data from: Estimation of regional annual abundance and evidence for increasing numbers of white sharks off central California [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.4022914
The matcher downloads the pretrained MiewID model
(conservationxlabs/miewid-msv2) on first use, and the demo checkpoint is
fine-tuned from it. Cite MiewID as its authors ask:
@misc{WildMe2023,
author = {Otarashvili, Lasha},
title = {MiewID},
year = {2023},
url = {https://github.com/WildMeOrg/wbia-plugin-miew-id},
doi = {10.5281/zenodo.13647526},
}