Case Study: Adobe Stock improves search relevance with Appen

How Adobe Stock used the Appen platform to build models to improve search relevance for their customers

Adobe, the global leader in creative software, uses Adobe Stock to manage a massive library of more than 200 million assets. The challenge was making every asset easy to discover, since uploader-provided metadata could be incomplete, overly broad, or inaccurate and did not always reflect how customers actually use images. Adobe worked with Appen and its platform to improve search relevance for Adobe Stock.

Appen helped Adobe create highly accurate training data by having annotators draw polygons around key image areas, such as copy space and object isolation, that are important for marketing use cases but not captured in standard metadata. This human-labeled data powered models that surface more useful assets faster, helping users find the right images without scrolling through pages of similar results. The result was a more effective search experience across Adobe Stock’s huge catalog, including hundreds of thousands of new uploads each day.

Adobe