Melanoma is one of the most severe type of skin cancer caused by the DNA mutations in pigment cells called melanocytes. The current diagnosing methods involves visually examining lesion images which makes it time consuming. This makes it an active area of research which is being pursued using various deep-learning models. However, classification models that are developed using limited labeled training data might have a negative impact on the diagnostic procedure. This study presents a self-supervised learning-based model trained on dermoscopic images for automated melanoma recognition, in which feature extraction is accomplished using unlabeled data, and further fine-tuning is done using labeled data for classification. Additionally, an intermediary step is included for augmentation using Deep Convolution Generative Adversarial Networks (DCGAN) to enhance the number of unlabeled images for the pretext learning step. The trained model improved the accuracy by 4% as compared to just training using the original dataset. The proposed method can be extended to a broader range of applications in the medical area, where there is a real scarcity of data to train models.

Adversarial attacks against deep learning models have gained significant attention and recent works have pro-posed explanations for the existence of adversarial examples and techniques to defend the models against these attacks. Attention in computer vision has been used to incorporate focused learning of important features and has led to improved accuracy. Recently, models with attention mechanisms have been proposed to enhance adversarial robustness. Following this context, this work aims at a general understanding of the impact of attention on adversarial robustness. This work presents a comparative study of adversarial robustness of non-attention and attention based image classification models trained on CIFAR-10, CIFAR-100 and Fashion MNIST datasets under the popular white box and black box attacks. The experimental results show that the robustness of attention based models may be dependent on the datasets used i.e. the number of classes involved in the classification. In contrast to the datasets with less number of classes, attention based models are observed to show better robustness towards classification.