AI services in cloud-based image management – what to look out for

The use of artificial intelligence (AI) in processing large volumes of images is very effective and has become increasingly important in recent years. Within a very short time, faces can be marked on thousands of photos, for example. And the material can then be grouped according to different people. Face recognition is of course only the impressive special case.

Via machine learning, which is basal to most forms of AI, all objects with similar features can in principle be identified if appropriate training has preceded. So it’s no surprise that AI services can now recognize and classify any everyday object in an image, and then generate appropriate metadata such as keywords (tags) or generic terms. Even collections with millions of photos can be made accessible within a day using conventional computing capacities.

Automatic color and object recognition, symbol image.

Processes that require language processing, can also be accelerated by AI. Read about it in our article ChatGPT in Media Management . One should also mention the use of AI technology in generating new images (see DALL-E, Stable Diffusion and the like).

Industry relies largely on major suppliers

It is no secret that in the digital age, Internet corporations and intelligence services use such technologies for their own purposes. It is equally obvious that cutting-edge technology is mainly developed within large corporations or well-funded research institutions. However, companies such as Microsoft and Google are not only promoting the development of artificial intelligence for their own use. Generating further revenue by licensing proprietary AI software to third parties is at least as important. Licensees are therefore business customers who would like to integrate AI algorithms into their own products. But they often shy away from the high costs and time required for in-house development of the same.

This is also true for most providers of digital asset management (DAM) systems. The use of AI in image management has been recognized and implemented as an important trend in Germany as well. But – and this is a problem – there is a widespread reliance on large providers that are based in the U.S. and are not subject to the data protection laws of the European Union.

Data sovereignty only with self-developed software

This data protection deficit is rarely disclosed voluntarily by providers. As a layman, you don’t necessarily know that software providers are increasingly relying on third-party components – true to the motto: Why reinvent the wheel? Preference is of course given to open source software. However, some innovations are not open source and can only be used for a license fee. In this context, one speaks of proprietary software.

To make a long story short. If large parts of a software rely on services that run via servers that are located in the USA at best, then data sovereignty is not in good shape. Again, think about automatic face recognition. Here we are talking about sensitive data, about biometric information that concerns facial symmetry and should not fall into the wrong hands. If you plan to process personal data with your image management system, you should always specifically address the issue of data sovereignty when choosing a provider. Ideally, you should choose a provider that guarantees the following:

• • Development of all software modules in-house and in Germany
  • Hosting exclusively in the European Union, preferably in Germany
  • Data protection compliant storage of biometric similarity vectors
  • Continuous further development, especially with regard to data security

Only if these points are contractually defined do you have legal certainty and can assume that your data will not leave the defined sphere of activity.

Be flexible with trainable AI solution

Flexibility is another important point to consider when purchasing AI-based image management software. What good is an AI platform if it only recognizes everyday objects? It should also identify specific products or individual logos. In other words:

Make sure you obtain software that allows you to train AI modules easily and at the same time with a high success rate.

Choose a transparent billing model

Last but not least, a sentence on the subject of pricing. Of course, you should assume that an innovative software solution with proprietary AI service that meets the highest data protection requirements has a certain price. As a rule, cloud-based solutions are billed via a SaaS licensing model. Here, you should make sure that the pricing model is also transparent in terms of AI functionality:

Are there any special costs for the use of the AI modules? If yes, is a flat fee charged or is billing proportional to usage? (For example, via an exact record of AI actions performed).

Test the AI platform of the teamnext | Media Hub

If we have aroused your curiosity and you would simply like to try out the various AI functions in the context of an image management or digital asset management system, then you can get started almost immediately with a free 14-day test phase for the teamnext | Media Hub and it’s integrated AI platform. In addition, you can of course book an appointment for a free online product demo with one of our experts at any time. Simply use our contact form for this purpose.