Recent deep learning studies have been explored in largely two directions: 1) Development of a strong model with efficiency. 2) Development of robust training methodologies to improve the target models. EfficientNet is a very successful model architecture which is investigated in various applications [1]. Based on EfficientNet, recently, various training methodologies have been investigated in [2] to improve the performance and the robustness by computer vision and machine learning communities. In this project, it is investigated the effectiveness of EfficientNet and robust training methods in the medical image analysis area. In particular, it is studied on finding the optimal data augmentation strategy using RandAugment for the target application (skin cancer diagnosis) [2]. Finally, the approach is evaluated on ISIC 2020 dataset and participated in MICCAI ISIC 2020 challenge [3].
[1] Tan, M. and Le, Q.V., 2019. Efficientnet: Rethinking model scaling for convolutional neural networks. ICML
[2] Cubuk, E.D., Zoph, B., Shlens, J. and Le, Q.V., 2019. Randaugment: Practical automated data augmentation with a reduced search space. arXiv preprint arXiv:1909.13719.