##Baseline:

For the sake of readability and due to many visualizations, the implementation of the traditional methods is saved in two separate notebooks for each dataset,namely cifar.ipynband monkey.ipynbrespectively. In addition, due to high computation time, for cifar dataset the mega histograms and kmeans models created using SIFT method are saved as mega_histogram_train_10.txt, kmeans_model_10.pkl, mega_histogram_train_50.txt and kmeans_model_50.pkl. Similarly for the monkeydataset, those files are saved as kmeans_model_10_monkey.pkl and mega_histogram_train_10_monkey.txt.

##More advanced methods:

tensorflow models:

• All trained models and histories are saved in the models directory for the datasets cifar and monkeys (both are dictionaries in the model dictionary)
• Each model starting with tf (tensorflow) can be started by opening the file. The nomenclature is tf_model-architecture_dataset.py. The script checks, if a trained model is available or will start training and predicting otherwise. Trained models can be downloaded from https://github.com/nadiaverdha/MachineLearningEx3/tree/main (copy the model folder to your cwd)
• Generally, the data must be in a "data" folder in the current working directory (os.getcwd()) following the initial download structure. import_cifar.py requires the described folder structure from the download from http://www.cs.toronto.edu/~kriz/cifar.html
• evaluate_nn_models starts an evaluation process of all models based on tensorflow.keras. If the models are not train it will result in an error. The provided run_model() function allows the display and save of train history and convolutionmatrix as well as a print of the model architecture.