The use of machine learning and artificial intelligence (AI) has received a lot of hype in the digital asset management (DAM) community, and rightfully so. One of the biggest pain points for DAM software users is finding time to tag their growing library of digital assets with accurate and clear metadata so they can easily search for and find their visual content.
With more and more DAM software providers rolling out image recognition and auto-tagging tools, we can’t help but stop and wonder – Is image recognition technology finally ready for DAM?
Let’s see what the DAM users think.
We asked our customer base to participate in an 11-question survey. Participants were asked to share their experience with four image recognition technologies, as well as their organization’s present and potential use of AI.
Forty-nine Widen customers from a variety of industries (manufacturing, technology, education, biotech and pharmaceuticals, financial, consumer goods and services, and government) completed the AI Image Recognition survey in August.
To ensure participants had adequate exposure to each image recognition technology and its keyword suggestion and auto-tagging capabilities, we gave them access to a test environment within the Widen Collective that we connected to the following image recognition services via each provider’s API:
Here’s what we learned!
Some image recognition technologies are better than others.
Participants were asked to rank, on a scale of 1-10, and comment on the helpfulness of the metadata tagged by each technology platform.
As a whole...
Participants felt that all four technologies tagged images with metadata that was frequently inaccurate and often too broad to be helpful with searching.
On a scale of 1-10, the highest weighted average rating among all four of the technology providers was a 5.27, suggesting that the auto-tagging technology is somewhat helpful but has significant room for improvement.
While weighted averages and participant comments clearly placed Clarifai as the frontrunner and Amazon Rekognition as the least helpful of the four that were tested, results were less clear for Cloud Sight and Imagga.
If you only consider ranking order, Imagga beat Cloud Sight; however, participant comments suggest that Imagga is superior. So, let’s call it a tie for the runner-up position.
Most DAM software users would find value in using an image recognition service but aren’t willing to invest just yet.
We asked participants, “Would you find value in using any of these image recognition services for auto-tagging your assets? Why or why not?”
Fifty-three percent of participants responded positively, indicating that they are excited to find any way whatsoever to improve search results and reduce the amount of time spent manually tagging their digital assets.
The remaining 47 percent of participants answered “no” or “not yet/maybe if the AI technology improves,” specifying that the service was still too general or inexact and that the time spent cleaning up inaccuracies outweighed any potential time savings.
Although the majority of respondents indicated that they would find value in an image recognition service, only 4 percent of these DAM software users said with certainty that their organization would invest in the service today.
Of the remaining respondents, 35 percent stated that their organization would not pay for auto-tagging functionality. The remaining 60 percent answered “I don’t know,” with many commenting that they would first need to consider the cost.
So, how can image recognition technology potentially help DAMsters?
The majority of digital asset management users believe that image recognition technology can potentially improve their efficiency and save them time and money.
Although only 17 percent of participants provided examples of how their organization is currently using AI or machine learning capabilities, respondents were quick to expound on the potential of such technologies.
When we asked participants how “image recognition technology could help enhance or improve your usage of DAM,” they identified a few areas of high value:
- Metadata/keywords: Standardize and fully or partially automate keyword creation, guide taxonomy development, assign or blacklist keywords from a controlled vocabulary, trigger new keyword ideas not previously considered, and supplement the manual creation of metadata across all images, including stock photos or user generated visual content.
- Related assets: Recognize similar images and automatically populate the same set of metadata across related assets, and easily locate similar digital assets according to style, composition, subject, or product type. Also, detect duplicate assets for easy deletion and locate related images outside of the DAM ecosystem, for example, on social media.
- Facial recognition: Identify people such as employees or spokespeople, and use machine learning for auto-tagging subsequent reocurrences of the same individual.
- Advanced search: Search by color, text within an image, or images and placeholder graphics in a video. Incorporate more search engine optimization (SEO), so results are prioritized based on ranking factors like query volume and relevance.
- Non-admin support: Maintain quality control in organizations that require internal and external content contributors, other than a designated DAM administrator, to upload and tag digital assets like user-generated photos, client creatives, or visual content from various departments.
In conclusion - The people have spoken!
The majority of DAM pros that participated in our survey feel that image recognition technology isn’t 100 percent where it needs to be yet, but they still see the value it can create in the immediate and near future.
At Widen, we’ve been exploring image recognition technology partners for over five years, and boy, have they improved! We've witnessed the unique strengths and weaknesses of each machine learning platform, and how an image recognition technology that is right for one organization might not be the best fit for another. We believe that a combination of human intelligence paired with the right AI technology, or combination of technologies, for your business can help drive efficiencies for today’s DAMster.
Tell us the DAM truth. What do you think?
Is image recognition technology ready for DAM?
Which image technology services do you think are the best?
How is your organization currently using AI and machine learning?