The Effect of Using Augmented Image in the Identification of Human Nail Abnormality using Yolo3

Human-nail abnormality manifests the status of nail’s health and human health in general. Terry’s nail is common to people with severe liver disease. Spoon nail can be found in people with diabetes and heart diseases. High cholesterol causes the Splinter Hemorrhage abnormality in nail. Although studies on Human-nail have been developed, there is still a lack of datasets to further the study on nail to serve as an additional tool for diagnostic purposes on specific abnormalities: Splinter Hemorrhages, Terry's nail, and Spoon nail. This study aims to determine the effect of using augmented images in training and testing nail image dataset to identify nail abnormality. The study compares three models: an unaugmented model, an on-the-fly model, and a manually augmented model using the open-source python image augmentation library imgaug, to identify Splinter Hemorrhage, Terry's nail, and Spoon nail abnormalities with its associated diseases using Yolov3 on a Raspberry Pi 4 model B with libraries like OpenCV, Keras, and TensorFlow. The manually augmented model achieved the highest accuracy of 91% which is 5.58% higher than the on-the-fly model and 13.92% higher accuracy than the unaugmented model..