One-Class Classifier for Real Fingerprints

Using the Generative Adversarial Network (GAN), I want to create three (3) separate one-class classifiers for classifying real fingerprints using :
1. The trained discriminator (classifier 1)
2. The trained generator (classifier 2)
3. Both the trained discriminator and generator (classifier 3)

Output (results) of each classifier should show:
1. The classification accuracy
2. A confusion matrix on test data (for each classifier)
3. A graph to compare the level of accuracy for all 3 classifiers created.
4. A 2D t-SNE visualization to display the live and fake fingerprints (for each classifier)

Datasets will be provided.

Also, project should be executed using the Kaggle.com platform.