VisageCloud: The scalable face recognition API


We launch VisageCloud, a full-fledged face recognition solution, based on state-of-the-art deep learning, neural network and computer vision algorithms. It is available both as cloud-based software-as-a-service and as an on-premise installation. In support of the key recognition feature, the solution has built-in face key point detection and alignment, face classification on several attributes, such as gender, age group, eye color, hair color, skin color.


To showcase the face search capability, VisageCloud has created a demo allowing any user to find the actors most similar to the people depicted in an uploaded picture. The entire process, containing face detection, face classification and face search runs in three seconds or less. The application searches through over 5000 known actors. The demo application is available at


Unlike similar offerings in the market, VisageCloud does not require storing or comparing picture. In a process that is more computationally-efficient and safeguards privacy, the original image being recognized is converted into a compact signature describing the face, known as a feature map or face hash. This allows for performing tens of thousands of face comparisons per second on a conventional computer, with possible speedups to hundreds of thousands of comparisons per second for mission-critical, real-time applications. Moreover, the original picture cannot be recovered from the compact signature, thus guaranteeing the anonymity of the user. By exposing an easy-to-use API, the solution can be leveraged in a wide variety of fields, from vending machines and info kiosks which can remember the preferences of regular customers, to smart surveillance systems and in-door digital advertising displays which can report analytics of the audience and adapt the advertisement in real-time.


The team behind VisageCloud is comprised of Bogdan Bocșe, cloud solutions architect, and Cosmin Nicula, software engineer.


“While we are very happy with the current results of both face recognition and face classification, we are focusing on improving and releasing new features in the near future. An example of such features include lateral face recognition and periocular face recognition, using just the area around the eyes. We are very excited to be working on this.” Bogdan Bocșe


”The current version of VisageCloud is the result of six months of research, development and testing. We realize that this is just the beginning, as we need to reach out the mobile, augmented reality and virtual reality markets. We also think our face recognition service can help retailers deliver on the promise of an omni-channel view of the customer. We are looking working closely with partners to better understand their needs, so as plan our next steps.” Cosmin Nicula
More details are available on

Bogdan Written by: