Image Processing and Deep Learning for Classification of Skin Lesion Image

The level of contrast and optical comparability’s between skin lesion images make skin disease classification a challenge. Skin lesion image classification can be great help for people in remote places. This research propose classification of skin lesion such as: Acne, Atopic, Psoriasis, Seborrheic and Ringworm, with the application of computer vision based application of deep learning such as Sequential model and ResNet-50 model. For pre-processing the image pre-processing technique such as SMOTE and Histogram Equalization (HE) is apply to balance the minimum classes of images in the dataset and to adjust the contrast of the RGB images. After experimentation, it is found that out of the two models, Sequential and ResNet 50, Sequential performed better and achieve accuracy score of $96 \%$ as compared to ${7 6 \%}$ by ResNet 50.