Real-Time Image Tagging for Real-Time Accuracy

Challenge

Identify celebrities pictured in client-provided images with rapid turn around. Manage spikes (e.g. related to large events like the Academy Awards) in incoming images so that tagging is nearly real-time.

Accurate image recognition algorithms have progressed rapidly in recent years, but human supervision is still key to maintaining accurate results. This provider of stock photography needed to confirm the identities of well-known figures in their photos. Mistakes would mean erroneous search results for their users, and unreliable datasets, so accuracy was fundamental.
Speed was also a big need for the client; during large events, there was often an influx of images needing tagging. This meant they needed a workforce that could expand in order to deliver on a quick turn-around.

Solution

iMerit teams worked daily, tagging images and enhancing automated identification results. As needed, teams ramped up hours and team size in response to image-heavy events such as The Academy Awards.

Results

iMerit tagged over 3 million images – often hitting over 5,000 images tagged in one day – with 98% accuracy. Teams improved productivity 150% from the beginning to the end of the engagement, enabling up-to-the minute, accurate image datasets for the client and their users. The client was continuously satisfied with iMerit’s work, and iMerit teams developed not only deep expertise, but also a fondness for the client and the work they did.

Client

Global Digital Media Company

Service

Annotation + Tagging, Categorization, Verification + Validation

Vertical

High Tech, eCommerce


RELATED CASES

Share this post with your networks:Share on FacebookTweet about this on TwitterShare on LinkedIn