Case Study: A Fortune 100 company achieves accurate isolated product images with Sama's human-in-the-loop image masking evaluation

How a Fortune 100 company improved its image masking algorithm to develop new, isolated product images

A Fortune 100 Company needed better image masking to produce isolated product images for advertising and engaged Sama to evaluate and validate generative model outputs using human-in-the-loop model evaluation and data annotation services.

Sama’s data experts reviewed original and masked image pairs against criteria for quality, segmentation, and comprehensibility, flagging errors and generating additional labeled training data to fine-tune the model. The result was improved model accuracy and fewer hallucinations, faster production of clean, isolated product images that boost ad performance, and an overall speed-to-production benefit consistent with Sama’s claims of getting models into production up to 3x faster.