Predictions made using a CNN-based image classification software. Given an input image, the model, available at https://doi.org/10.5281/zenodo.5091813, is able to predict values for five properties, namely production 'timespan', 'production place', 'technique', 'material' and 'depiction'. It has been trained based on a May 2021 snapshot of the Knowledge Graph. The multi-task learning (MTL) variant is being used in a multi-class classification (mutually exclusive classes) fashion based on the softmax function for computing the class scores.