Hype or Game Change: Our CEO David Backstein talks about AI in DAM

Photo by Jonathan Farber on Unsplash

How can AI help DAM users today and how can it be a game changer in the future? In the latest episode of the DAM Evangelist podcast our CEO David Backstein and DAM expert Ulrich Leidl address many important issues that drive DAM users when it comes to AI.

Generic tagging solutions vs. custom trained neural networks

Currently, automated tagging solutions implemented in DAM offer only generic keywords. This may be sufficient as an initial tagging approach, but most DAM users have very specific assets where deeper tagging is required. An example: For a bicycle store it is not sufficient to tag an image with “bicycle”. Instead more keywords related to this industry are required to describe the image appropriately and to make it retrievable (e.g. brand, type, size, color). AI can help here with custom trained neural networks.

These neural networks have been trained with millions of images from a specific subject area with respect to specific features. The networks can therefore classify and tag new data of the same subject area very well. The big hurdle is that a lot of training data is needed to train the AI and get good results. However, today there are also methods where it is possible to achieve good results with less training data.

It’s important to understand that those solutions never work out of the box and that AI is not self-learning per se. It’s always an individual solution trained for an individual image set.

A glimpse of the future: DAM with evolving keywording system

Visual Search AI. High performance image search for your website or app.