Text this: Detection of Adversarial Attacks Using Deep Learning and Features Extracted From Interpretability Methods in Industrial Scenarios