ArtiVision is a convolutional model trained on the WikiArt dataset. The primary goal is to classify images by their artistic style and visualize model results and behavior.
The dataset used in this project can be downloaded from Kaggle: WikiArt Dataset. It contains 176436 images and a csv mapping each to their respective info.
To preprocess the data, run the files in data_preprocessing in the order group_by_style.py then split_style_data_to_train_test.py.
Run main.py to train the model with choosen arguments. Edit hyperparameters.py based on machine capabilities. Model architecture is defined in models.py.
Visualization and various testing scripts can be found in the results_viz folder. To run top-n prediction accuracy, run top5_accuracy.py then either top5_correct_analysis.py or top5_dr_analysis.py.