The computerized detection of hair type and style sounds really interesting. How far down this path are you now, are you planning on a deep-neural-net (e.g. CNN's) or something simpler?
Would certainly make for a unique image-classification problem, with its own interesting challenges; a collection of images of black women with different hairstyles will be a lot more visually similar than a collection of dogs/cars/mountains.
Yup, it's interesting! We're in the early stage of this, currently experimenting with [Keras](https://keras.io/) and focusing on the inception pre-trained network. Our goal is to retrain the top layer of the network using the tags/hair types/categories we're collecting on the app as labels.
Would certainly make for a unique image-classification problem, with its own interesting challenges; a collection of images of black women with different hairstyles will be a lot more visually similar than a collection of dogs/cars/mountains.