Pre-trained model for classification of images of pollen grains, for use in Scivision.
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.
The python code itself is in the directory pollen_classification_model/
__init__.pyjust allows this directory to be treated as a python package, and imports the contents ofmodel.pyinto the package's namespace.model.pythis 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 apredictmethod. It should also have a way of loading the weights.