/honeybee_species_model

try adding a trained model to the SciVision catalogue

Primary LanguagePythonMIT LicenseMIT

honeybee_species_model

Repository containing a pre-trained computer vision model for identifying species of honeybee, to be included in the Scivision catalog.

Installation

To install this package into your environment, run pip install . from this directory. Note for MacOS users with new M1/M2 machines: Tensorflow doesn't seem to be installable from pip for ARM-native python. Can either use Intel-compiled python (e.g. via anaconda), or install tensorflow via conda. If the latter, you may need to remove tensorflow from requirements.txt to avoid problems when running load_pretrained_model in Scivision.

Contents of this package

The python code itself is in the directory honeybee_species_model/

  • __init__.py just allows this directory to be treated as a python package, and imports the contents of model.py into the package's namespace.
  • model.py this is the wrapper for the model itself. It ensures that images on which the model will be run are the correct size, and it provides the interface that Scivision expects. It should contain a single class, which has a predict method. It should also have a way of loading the weights.