Skin Disease Classification Using Deep Convolutional Neural Network on Jetson Nano Developer Kit

Skin cancer is one of the cancers prevalent in world health. They are the cancers that occur in the top most layers called epidermis part of skin. The main cause of skin cancer is due to excessive exposure of UV radiation. A cancer can be benign or malignant. Many deep learning algorithms are proposed in literature to classify skin lesion images. However, there is not much work has been done to implement the models in hardware. In this work, a convolutional neural network (CNN) based model is used to train dermoscopic images of PH2 dataset and implemented on jetson nano board to classify the images as malignant or benign. Further, the proposed model is compared with two pre-trained models namely MobileNet and ResNet50 and found that proposed model outperformed than these two models. The proposed model detects the melanoma skin cancer with 93.33% accuracy on test set and hence can be used as second opinion to dermatologist.