Classification and Grading of Auto-Examined Acne Face Images using Low Complexity Support Vector Machines

Acne vulgaris, frequent problem encountered by huge population many people’s especially in puberty around the world. Acne images can be pre-classified using machine learning model like SVM and graded accordingly to variable acne severity. Acne vulgaris is a most common acne type which has always been a key interestdomain for researchers and academia. There is a lack of dataset in this domain since the personal images can not be shared without individual choice. Thus a data augmentation is applied on a small dataset to generate a large dataset to achieve better accuracy. In this paper, Acne face image classification is performed using SVM supervised machine learning model. The optimum parameters are found using the grid search for radial basis kernel and classification score of different kernels is compared. Acne grading is done as mild, moderate and severe based on the severity of acne. The result is validated in terms of precision, accuracy and complexity and it is found to be effective to be used in low resource devices.