Seedling FiftyOne

An introductory approach to using a FiftyOne dataset interface with seedling images using BioCLIP 2 classification attempts is available in seedling-fo.py.

To set up:

• Create and activate an environment.
• Install libraries in seedling-fo_requirements.txt with e.g.:

uv pip install -r seedling-fo_requirements.txt

To run:

python seedling-fo.py <path-to-images-dir>

Additional flags exist as well. Currently, you may:

• specify a desired --confidence level for open-ended prediction
• use a --device (e.g. cuda)
• use --custom-labels pointint to a CSV with a column of custom labels to predict among
• specify a --dataset-name if one has already been executed
• --force-reprocess to remove dataset contents with newly processed data