How media organizations and content libraries are hosting computer vision solutions on premise to regain control and accelerate commercialization.
The last few years have challenged a lot of our ideas regarding ‘normal’ things. A visual representation of an ‘office’ is no longer restricted to desks, chairs, conference rooms and formal-wear; the concept of a ‘home-office’, that became popular during the novel COVID pandemic is now widely prevalent, as is the need of recognizing the relevant visuals that go with new concepts like these. New political and socio-cultural events have given rise to thousands of unique, innovative imagery that symbolize and encapsulate their messages.
With news cycles moving at a mind-boggling pace, photo libraries that can return relevant images for these social phenomena are in a much more profitable position than the ones that cannot. Consequently, an increasing number of them are choosing computer vision solutions to automate the classification of the millions of visuals being produced everyday. Improving search matches even by the tiniest percentage can make the difference between commercial success and failure.
However, in order to accurately recognize visuals, computer vision tools require substantial processing power to manage the visual archives consisting of millions of images. This typically requires support from one of the big cloud providers such as Amazon, Microsoft and Google who offer machine learning training and classification based on the media uploaded to their networks.
The Tech-Cloud is not a fluffy place between rainbows
Cloud providers often negotiate their prices based on extracting metadata from the uploaded media. This metadata is then used to train their own machine learning algorithms. Many photo libraries, that have spent decades building up their archives, are understandably reluctant to give this kind of proprietary information to Big Tech. Moreover, the cloud service depends on data transfer between the cloud, your system and clients. Therefore, the quality of connection with a cloud provider does not always guarantee a consistent, split second response to a visual search query. Even slight delays or breaks in service may risk both income and long-term customer loyalty.
The most significant concern that rises with regard to using cloud services is that of data privacy and data protection. Uploading visuals to the cloud often entails handing over sensitive data to a third party. There are several rules that govern such sharing of data. GDPR (General Data Protection Regulation) and other legislation that are in place to ensure that no data is stolen or misused. These laws also make it increasingly expensive and complex for organizations to exchange sensitive data.
Live on the 'Edge' instead
‘Edge computing’ is a computer vision model that works on your own systems. Proprietary data and other content never leaves the organizations’ servers, whether they are hosted on premise or with a co-hosting partner. Media platforms are turning to this solution to retain better control over their archives and reduce their dependence on third party servers. The solution is delivered in the form of a Software Development Kit (SDK), which includes compressed algorithms that run in the background of the existing media library or digital asset management tool.
Superhuman Vision™, Mobius Labs’ next-generation edge solution, is pre-loaded with thousands of tags, including abstract concepts, emotions, and actions. In addition to facial recognition and video classification, the technology includes aesthetic ranking models which identify the best content in a collection. Apart from ‘out of the box’ tagging, the SDK enables non-technical users to train the software and create custom tags based on a relatively small number of images. Media platforms and photo agencies can thus keep ahead in the race of finding visuals that suit the ever-evolving and rapid news cycles. With technologies like that of Mobius Labs’, it becomes convenient for them to also re-classify content that reflects changing socio-cultural attitudes regarding diversity and the environment.
The software enables photo libraries to continually cater to the needs of their clients, and simultaneously helps them branch out to a wider audience to increase revenues. More significantly, they can do so while retaining complete control over their content and data. Superhuman Vision™ is deployed as an SDK that integrates seamlessly with client systems, thus leaving their data where it belongs- with them.
Author: Dr. Appu Shaji, CEO and Co-Founder, Mobius Labs.
Editor: Trisha Mandal, Content and Comms Werkstudent, Mobius Labs.