Implement evaluation stuff

[muellerdo]

• Training Evaluation via csv_logger or keras history
• Performance evaluation with preds matrix from predict() and label_ohe
• Compare prediction function (pred_list=[pred_a, pred_b, pred_c], labels_ohe)
• dataset analysis (input_interface)
• model complexity

Training Evaluation

• epoch vs loss - all in one plot
• epoch vs loss - facet_grid for each model (with all metrics)
• epoch vs loss - smoothed over models

Performance evaluation

• ROC plot
• barplots with all kinds metrics
• heatmap for confusion matrix
• CSV table with metric computations

Compare prediction function

• facet_grid large barplot dataset vs metric
• Performance gain

Dataset analysis

• class distribution -> barplot/pieplot & heatmap for multilabel (samples vs class)
• table with class distribution
• sample showcase -> visualize 4 randomly collected images from each class
• sampling?

[muellerdo]

Current Status:

• [x] Training Evaluation
• [x] Performance Evaluation
• [x] Performance Comparison between models
• [x] Dataset Analysis

[muellerdo]

• [x] Evaluation submodule Description