By David Alejandro Pabon, – The person identification problem is one of the most extensively researched problems over the past decades and has recently received an immense amount of commercial interest due to its increased reliability in comparison to past decades. The main reason for its increased reliability is the accelerated advance in both, deep learning techniques and hardware for AI processing that enable faster and more precise inferences. All of these are framed with the open-source community of developers, which makes it possible for new enthusiasts to start from a very upper-level starting point. The application of facial recognition and person identification on real-world surveillance systems involve technical difficulties such as dealing with noise and low-resolution features, combined with unconstrained poses, ambient occlusion, and cluttering. Therefore, sometimes it is useful to combine naïve face recognition models with robust key-point extractors and trackers backboned on clever deep learning architectures.
The state-of-the-art models that perform face recognition over high-resolution images are usually ill-conditioned for performing the same task over native low-quality surveillance face recognition tasks. And is very difficult to synthesize the training and testing data in a way that reflects how good the recognition systems are going to work over the actual surveillance system.
Surveillance of people is not something new, but it has been drastically enhanced by AI. When combining huge amounts of data with powerful mechanisms driven by AI, a person identification surveillance system can interfere with privacy and put in jeopardy sensitive data. But despite these ongoing challenges and ethical aspects, it is possible to say that both facial recognition and person identification have reached a point in which most of the issues are solved and there’s an increasing interest on marketing commercial systems that involve facial recognition and person identification and the companies are more interested in quickly ensure scalability, cost-efficiency, and user interfacing.
And most of these deploying challenges have been recently addressed by platforms that aim to provide efficient and cost-friendly services that empower high computing capabilities and APIs to individuals and small companies that aim to lead the way.
Every day more and more companies are starting to take a step front into the race of getting the most of advances on deep learning-powered computer vision and we are right on track.