We're looking at developing a machine learning algorithm (CNN) to assess whether the ID's that users upload are photoshopped.
We have developed an identity verification solution. It has several tests - the last one being a photoshop test. We have a manual review team that will do the first 1000-5000 images manually, which we will use as our input data into a convolutional neural network (CNN). Our goal is to test whether the images have been modified in any way, and to get a 95%+ success rate.
We're looking for a developer who has experience in CNNs, coding in Java.
We'll look to start the project in the third week of March.