That’s not how this works. You’re claiming an incredible breakthrough: extremely high facial recognition accuracy using 2D input in an adversarial setting. Extraordinary claims require evidence.
As I've mentioned earlier, we will be publishing in-depth results soon. Meanwhile, you can read here the review of public datasets https://medium.com/xix-ai/quantifying-the-inherent-bias-in-m...
Overview on adversarial examples is below. The upcoming post will cover the approach, performance and comparison. Stay tuned
Will this upcoming post address the storage and privacy concerns that were the main point of the GP?