Automated Assessment of Excessive Inbreeding of Dogs based on Deep Learning

Brachycephalic obstructive airway syndrome (BOAS) is a common disease that effects animals such as bulldogs and Persian cats,resulting from excessive inbreeding to produce an idealized pet. The physical manifestations of BOAS are a flat to concave face, smaller nostrils and excessive wrinkles. Unfortunately, BOAS has a range of negative impacts on the dogs,causing some problems themselves including skin irritation, breathing difficulties and a reduced lifespan. In this paper we explore the extent to which BOAS can be automatically quantified through deep learning. A range of deep learning models were explored including MLP, CNN, VGG16, VGG16 fine-tuning (later 3 layers), Xception, and Xception fine-tuning (later 33 layers) on a data set that was established for the work. Experiences and results of the application of the different deep learning approaches are presented.