– Case Study –
Continuous Learning For Facial Recognition In Images
A Leading AI Photography Platform requires continuous model retraining to effectively curate photo albums for their customers
Challenge
A leading AI photography platform delivers curated albums based on who is in the photo. Photographers upload large batches of photos that our client’s model matches with selfies. Model accuracy depends on correctly identifying multiple attributes that impact the success of model training and prediction. Use cases include people at concerts, corporate events, weddings, sports, etc.
99% Model accuracy
Results
SOLUTION
Our on-demand, flexible, and scalable data annotation team delivers near real-time judgments on the accuracy of the model’s predictions. Annotators must distinguish key features and vote on whether they match the model generated recommendation. Each photo is initially voted on by two annotators. Photos that are disagreed upon are sent to an experienced third annotator to break the tie. We continuously monitor arbitration rates to mitigate bias and cost overruns.
IMPACT
This AI photography platform is revolutionizing the way people share photos. The model saves time shortlisting images for both the photographers and end users. Our swift delivery and precision directly impact the user experience of their customers. This is an ongoing project for which we have already processed 2.5 million images in the first 20 weeks. Thanks to our services, their model has improved to 99% accuracy.
Meet an Expert
Our Team of Data Experts
A team comprised of data experts with extensive experience in developing AI-based data solutions for clients. Book a time that works for you and let us help develop a custom solution for your unique needs.