Automated Deep Learning Based Approach for Albinism Detection

As we all know, Albinism is a condition in which a person is born with a deficiency in melanin pigment. The colour (shade) of skin, hair, and eyes is a symptom of having a molecule called melanin. Pale skin, hair, and eyes are signs of Albinism. People with Albinism are more likely to get skin cancer, and their skin is highly prone to sunburn. In our dataset, which contains data on people with or without Albinism, deep learning frameworks like inception V3, VGG 19, VGG 16, and many more were used. In this paper, we have employed different hybrid combinations of pre-trained deep learning architecture for deep feature extraction in combination with different hand-crafted based classifiers for classification. After a detailed analysis of the various hybrid architectures, the best one was proposed that outperformed the other state-of-the-art architectures. We further performed accuracy assessment on several statistical measures such as AUC, accuracy, precision, recall and others. Our proposed hybrid architecture was compared with other architectures using ROC (receiver operating characteristic curve). Our approach produced an accuracy of 99.8% compared to other models.