pythonmachine-learningcomputer-visionimage-recognitionviola-jones

AdaBoost and Viola Jones: What training set to use?


I have implemented my own version of Viola Jones face recognition algorithm that uses AdaBoost as a meta for building a classification committee. My aim is to build a classifier that can recognize whether there is a human face in an image. I am struggling to find adequate set of training data to try out the algorithm. In particular I dont know where to find a set of negative images (i.e. images that do not contain a face). For the positive dataset I was going to try the Labeled Faces in the Wild dataset link.

What would be a good negative dataset?


Solution

  • Some solutions that might work for your problem are:

    Care should be taken when selecting this dataset, as it can easily introduce bias if you're not careful.