I am trying to build an experimental image classifier for the Mallampati score, used in anesthesiology, as a free community project. Basically, an input picture of the oral cavity with all amounts of scale, translation, and rotation variance would be presented, and the classifier would grade it as Mallampati 1, 2, 3, or 4.
- Do you think such a task is achievable?
- What would be a suitable image classifier algorithm?
- Would you think of any special requirement for the input picture, such as uniform size, file format...
Given the amount of shape, scale, translation, and rotation variance I will get in the input pictures, I was thinking that a multilayer perceptron neural network would be best. However, I have also looked into things such as LibSVM, thinking SVM could be an option too...
I am currently building a photobooth website as an attempt to gather some data. If you think the project is a good idea, and you would be interested in participating, drop me a line through raoul dot schorer at gmail.com
Those are the kind of pictures I would get (sorry for this pic, now you know what your dentist feels like :-)