Picturepark specializes in content management solutions for brands and agencies across the world. Their mission is to help organizations effectively connect, manage, and route content from its points of creation to where it’s most needed. However, manually managing the massive amounts of diverse media most companies now have can be unproductive and unwieldy. As such, Picturepark turned to artificial intelligence, in the form of computer vision (CV).
Computer Vision & Content Management
We recently spoke with Matthew Jones, Marketing Manager at Picturepark, to discuss the impact CV AI is having on the content or data-asset management (DAM) world.
One of the key impacts he noted was lessening the complexity around content management for organizations, especially ones with multiple offices and departments containing many large teams. Not only is there a staggering amount of content needing to be stored, but as much of it is diverse in nature, you have a pretty daunting task.
“Correctly categorizing and tagging media can be a highly time intensive process,” Jones said. “Users ‘see’ different things in visual content. People project their own knowledge and biases on to what they see. I’m not just talking about the ‘What colour is this dress?’ phenomenon of 2015. People think about images in all kinds of different ways. What one person sees as a Yorkshire terrier, another may see as just a small dog.”
This is why Picturepark ended up utilizing Clarifai’s AI to build a custom connector that let their customers “auto-tag” their visual media within the Picturepark environment they’re familiar with.
“While of course nothing sees with true objectivity, after all Clarifai is programmed by humans too, the wide library of available terms on Clarifai means that users can tag in many ways at once,” he said. “Once tagged, other users can then search in a variety of different styles: reaching the content they need, as fast as possible.”
A key challenge for the Picturepark team was ensuring that customers always felt like they were in control.
“Whilst we’re always impressed with Clarifai’s results, we also know how important it is for our customers to be the ones making all the decisions,” Jones said. “We got around this challenge by framing Clarifai’s results as suggestions, so the user decides exactly which keywords get added and to which content.
“We recently covered this challenge in a blog post all about auto-tagging, using the example of a clean or eco-friendly power station that sources its power from geothermal energy. Harnessing geothermal energy creates large amounts of steam but of course AI algorithms are likely to perceive an image this process as ‘smoke’ or ‘pollution’, after all - it looks very similar!
This is why we have to leave it to our users to choose how to best direct the power of Clarifai.”
And the team is happy with their choice of Clarifai to enable their users to take advance of computer vision technology.
“We’ve had a great experience with the Clarifai team,” Jones said. “Right from the start we were impressed with how responsive and helpful the staff were. We could tell from the word ‘go’ that Clarifai would make a good company to collaborate with, when we saw the amount of clear documentation that accompanied the API. This made building our first ever proof-of-concept a breeze and we haven’t looked back since!”
Proving the ROI of Computer Vision
Of course, before investing in a technology, you have to be able to prove its value. We recently launched a free ROI kit for how much a company could save by investing in computer vision for moderation., For Picturepark, though, it’s all about the value gained by customers.
“We’ve been in the content management sector for over two decades, and one observation we’ve made is clear: automation is a key area for our customers,” Jones said. “Content libraries are rapidly increasing in size, especially when it comes to visual content as this area continues to grow and grow, but staffing-up is limited. Different clients have different feedback, depending on how they use Clarifai but one thing is very clear: every one of them has recorded a high level of cost saving. This increases the more customers are willing to invest into a sound information architecture first: without doing this, even with the effects of AI, it would not be very sustainable in the long term.
“A major ROI for our customers is the amount of time they’ve saved and have been able to invest in things their labour is better spent on, other than manually tagging visual content,” he said. “An aspect we really enjoy here at Picturepark is seeing these results happen for our customers so quickly in the short-term. The way our customers see the return over the longer-term is through the multiplication of this factor - every month or year, their content library extends even further and the amount of time saved doubles, triples, and so on!”
The Future of Computer Vision and Content Management
To cater to the ever-growing content ecosystems of companies, the future of content management will need to be more automated and omnichannel.
“Content is becoming increasingly fileless, consisting of various pieces of information that need to be kept together in a as much of a channel-neutral master format as possible,” he said. “What’s key is that this content can then be used across the many different platforms that we use, such as mobile apps, websites, and other platforms. This is reflected in the design of our new Picturepark Content Platform, which we’re very excited to release.”
And when it comes to the future of implementing computer vision, Picturepark is excited to utilize a custom model through Clarifai.
“After talking about how different people ‘see’ in different ways earlier, this model seems particularly apt as it gives users the power to train a model to see in the same way as them,” Jones said. “We have many customers that have many specific products. We are excited in the potential of a Clarifai model that could ‘see’ in the same way as one of their product specialists.”