• abele
• adversarial autoencoder
• explainable artificial intelligence
• generative models
• image classifier
• psoriasis

In recent years, the exponential growth of machine learning and deep learning has reshaped numerous domains, ranging from computer vision and natural language processing to healthcare and finance. This expansion is characterized by increasingly sophisticated models, which have achieved extremely high performances, in many cases outperforming human beings. However, these models often operate as black boxes, lacking transparency in their decision-making processes. The research field of Explainable Artificial Intelligence (XAI) has developed to overcome these shortcomings, with the aim to enhance transparency, interpretability and trust- worthy in machine learning and deep learning models. This thesis explores the application of an Explainable AI system, called ABELE, to a real case scenario, namely an image classifier targeting psoriasis lesion severity, developed by siHealth Photonics srl. ABELE provides visual explanations to the predictions of the deep model, through examples and counter-examples which are respectively newly generated artificial images picturing psoriasis lesions with the same and different severity. During the project we implement and train ABELE using the dataset of psoriasis images made available by siHealth Photonics. We tested several model configuration and different training strategies finding that the artificial psoriasis images generated are of poor quality mainly due to the reduced size of the dataset at disposal and the high complexity of the images.