Dreamstime, has announced the implementation of a proprietary artificial intelligence (AI) system that uses sophisticated algorithms to screen submitted images. The machine learning tool is designed to examine how human editors at Dreamstime review images, and then adjust its parameters to best match the editors' various criteria.
The AI tool from Dreamstime will go beyond current image recognition software, as it can unite metadata with the context of the image itself. It can be applied to multiple areas of the submission process, for example notifying users they need a model release or recognizing a copyright logo resides in an image. The tool has the capability to detect features in submitted images that trigger rejections, and uses this same intelligence to label and localize each photo.
This is a next-generation enhancement that can not only recognize the subject of a photo, but also relate its compositional and commercial value.
Sources tell me that this announcement may be a little premature. Currently, Dreamstime has one software engineer working on the project. In addition, they have just asked remote reviewers for unpaid assistance in testing the algorithm’s results. It seems Dreamstime hopes reviewers will be happy to train their future replacement.
Not a full replacement, but once the algorithm is functioning smoothly it will cull the number of images reviewers must consider. Reviewers are currently paid based on the number of images they are asked to review.
It is also unclear, at least at this point, how many false results the algorithm is likely to generate requiring image creators to spend more time re-submitting images. Given the current stage of development it may be some time before the algorithm provides somewhat decent results unless Dreamstime hires a lot more coders.
Dreamstime says it will continue to utilize its team of human editors who understand the nuances of reviewing in terms of the legal, aesthetic, and technical aspects of an image. This team's ongoing work will be integrated within the AI system, allowing it to adopt some of the human element of the review process.
In addition, Dreamstime says both photographers/contributors and end users will benefit from the AI solution. Photographers will receive a streamlined review period for images, so they can have more content accepted and more quickly earn revenue for their work. Dreamstime estimates it will reduce the review period with 30-60% and in some cases, the review will be instant. End buyers/designers will gain from a wider selection of image content and a refined search capability that accounts for the AI-produced data.
"Constant innovation is a primary reason for our continued success in the stock image industry," said Serban Enache, the company's CEO and co-founder. "With our AI solution, we have created a continually improving tool that will conduct instant image reviews, which will streamline our approval process. And we can layer on other types of metadata points such as customer behavior to build a truly intelligent solution that helps us improve the end-user experience."
Among current functions we can include a search engine that helps editors determine if certain images were uploaded before and are already online, thus avoiding duplicating content, automatic recognition of certain landmarks that can only be sold as editorial licenses, automatic recognition of illustrations.
Future planned enhancements to the AI's functionality include the addition of customer behavior data, so the system can effectively "learn" what types of images are selling, and which ones are less appealing. It will do this by learning from various sources and combine that information with data from other previous steps to develop an accurate gauge of an image's likely value.