version number: 0.0.3 author: LUH
This software provides python functions for the classification of images and training and evaluation of classification models. It consists of five main parts: 1. The creation of a dataset, 2. the training of a new classifier, 3. the evaluation of an existing classifier, 4. the classification of images using an existing classifier and 5. the combined training and evaluation in a five-fold cross validation.
All functions take explicit parameter settings as an input and generally write their results in specified paths. A documentation of the functions' parameters can be found in documentation and further details are described in the SILKNOW Deliverable D4.6.
To install clone the repo:
$ git clone https://github.com/silknow/image-classification
$ cd ./image-classification
$ pip install --upgrade .
A pre-trained model that was created using this software can be download from https://doi.org/10.5281/zenodo.5091813. The training of that model is based on the focal loss using a mutli-task CNN architecture.
The user can download the provided classification model and directly start to classify new images by means of the function silknow_image_classification.classify_images_parameter.
Alternatively, the user can train an own image classification model using the provided software for a subsequent image classification. Therefore, example calls of all functions are provided in main.py using the provided data files. These function calls will perform all steps listed in the overview above using the provided knowledge graph export.