Will AI transform the digital asset management industry?

Martin Wilson says artificial intelligence will have a huge impact on the future of digital asset management.

martin wilson 180Are you aware that AI (artificial intelligence) applications are already used in almost every industry? If not that’s probably because in popular culture for a system to be artificially intelligent it needs to be able to ‘think’, like we humans do. Many computer scientists prefer the term machine learning for exactly this reason.

Often you won’t read about ‘artificial intelligence’ in a company’s marketing blurb, even if their products use machine-learning technologies. They may be conscious of how high expectations of AI have been in the past or perhaps they simply don’t think of the technology in those terms.

The same cannot be said in the digital asset management (DAM) industry, which has quickly spotted the potential of visual recognition APIs such as Google Cloud Vision to save huge amounts of time during the tagging process. Many DAM vendors are now announcing AI-powered functionality in their products.

Why are they so excited, and is there any substance behind the hype?

 

Metadata is important

The core functionality of DAM systems hasn’t changed much over the last 15 years. Fundamentally a DAM application is still a database that stores digital files and their associated metadata. The metadata (for example a title, description or keywords) plays a crucial role as it enables users to find assets easily. This is essential for any organisation with more than a handful of assets. At present, almost all of the metadata is entered manually by users during or shortly after upload.

 

A typical DAM system

When DAM systems first came out, over 15 years ago, on average it took an experienced user about 30 seconds to enter the metadata for a single image. Since then user interfaces have improved and technology has become quicker, but entering metadata still remains one of the most time-consuming activities when it comes to organising assets. It’s no surprise then that DAM vendors have been keeping a sharp eye on the growth of visual recognition technology. If keywords and even descriptions can be added automatically, saving 30 seconds per image for an organisation that owns millions of assets, this adds up to big savings.

 

APIs at present

Currently the available visual recognition APIs provide computer-generated tags, which vary in quality and are not always reliable. Where results can be impressive for generic subjects, they are not so good for more complex images. So, an organisation that uses a lot of generic images, like beaches, landscapes and so forth, could see a real benefit from using this technology.

Related: Patrick D’souza on the API economy – speed, creativity and saving lives »

Others, especially those with domain-specific images, might spend more time correcting keywords than they save.

One thing to remember is that the DAM solutions will almost certainly be using a third-party API to provide the actual visual recognition functionality, so neither the client nor the DAM vendor has much control over the quality of the auto-suggested tags. In addition, none of these API providers yet offer much in the way of customisation, which can cause problems, for example when terminology differs. A UK company may use the keyword ‘tap’ but the US associates would recognise the same image as a ‘faucet’, thus the results may not be universally suitable.

 

Small steps forward

While they wait for the image recognition APIs to improve, in terms of accuracy and opportunities for customisation, some DAM vendors are looking for creative ways to extract the most value from these technologies. For example, by combining visual recognition technology and human oversight in a user interface, it is possible to speed up tagging dramatically. This approach can be used to do things like create smart groupings of images.

In its current state, visual recognition technology may add value to your company’s DAM solution, especially if you deal mostly with large volumes of images containing generic subjects.

The suggested keywords will not be 100% accurate, but they may be good enough, especially when combined with manual functionality.

In general though, it is likely that there won’t be a large adoption of these technologies until the results improve. But this is an area that is developing quickly, and the DAM solutions that are supporting this functionality now are well placed to quickly leverage the potential of AI as the APIs get better.

AI technology is an emerging and fast evolving field that will have a huge impact on DAM in the future. That future may be some way off, but when it arrives and we no longer have to manually add metadata to assets, the upload process in a DAM application is going to change forever.

 

Martin Wilson is the founder of Asset Bank.

 

* * * * *

Purchase a subscription to Marketing

* * * * *