Our client is working with a car manufacturer to develop machine vision technologies that can track and understand human facial expressions. In order to develop this technology, our client needs large image data sets containing accurately tagged and annotated images.
Our team created a custom interface where the image is loaded and an iMerit team member marks approximately 50 unique facial landmarks using a markup tool. The interface works seamlessly with our quality control processes, enabling quality control teams to not only review, but also tweak points as necessary.
Photos of 125 different people totaling over 90,000 unique images. Of these, 50 percent were annotated with over 40 facial landmarks, others were categorized into pre-defined rejection categories. Project is ongoing and growing.