Automating the discovery of nude pictures has been a major problem in computer vision for over two decades now and, because of it’s rich history and straightforward goal, serves as a great example of how the field has evolved. We’ll use the problem of nudity detection to illustrate how training modern convolutional neural networks differs from research done in the past. In particular, we'll discuss the notion of the deconvolutional network and demonstrate how it can be used to visualize intermediate feature layers.
Ryan Compton is a data scientist at Clarifai. He was previously on the research staff at Howard Hughes Laboratories and holds a mathematics PhD from UCLA.