As you may be aware, Paravision was born out of a face recognition toolset developed for a photo and video storage and organization service called Ever. As the Ever service was becoming popular, one of the most requested features was to automatically organize photos of the same person. It was clear that face recognition—as used by many major photo storage services—was the right solution to this problem, but when the team at the time found no suitably accurate models in the market, it embarked on creating its own. The internally developed face recognition feature was very well liked by Ever users, and became a key part of the Ever service.
At one point, Ever was a high-growth business, serving millions of active users around the world. However, over time it became increasingly difficult to compete with the market, product, technology, and ecosystem dominance of major tech companies who offered a competitive photo storage capability. In the face of these conditions, and an increasingly unprofitable consumer business unit, it became clear that the Ever service was not going to be a long-term, viable business. We made the decision to close the Ever service this summer, announcing it to our customers on July 31st.
While the Ever service was unable to compete in the consumer space, our face recognition and related computer vision capabilities thrived as an enterprise-oriented business unit. Due to our technical leadership in Deep Learning and AI, we have been able to out-compete biometrics and computer vision providers of all sizes and fortunate enough to partner with some of the best solution providers in the market, helping to transform building security, access control, payments, travel, and government services.
While we are disappointed that we had to close the Ever service, this is a positive inflection point for Paravision face recognition, as it allows us to address and allay concerns and confusion about the use of Ever user data for training enterprise AI models. Along with the closure of the Ever service, we will be deleting all photos and videos from Ever and will not use them to train Paravision face recognition models.
We want to be clear on some critical points:
- We never tied user identities to our enterprise face recognition products, never exposed user identities or data to enterprise customers in any way, and have never sold user data to any party at any time. At the same time, we recognize the confusion it caused to use this data to train and improve models for both our consumer and enterprise customers, and we’re happy to move forward with this confusion behind us.
- Again, going forward, we will not train any models on Ever user data.
While concerns have been expressed about the use of Ever data for Paravision face recognition, it is in fact our strength in Deep Learning and AI that has enabled us to create the most accurate face recognition toolset available from any company in the US, UK, and Europe. We look forward with excitement, knowing that we will continue to offer performance in line with what the market and our partners have come to expect from us—the best accuracy, the most robust performance in the widest array of challenging conditions, and the broadest range of deployment options—without any of the confusion that may have previously come along with it.
-The Paravision Leadership Team