Case Study: Google helps blind runners run independently with Sama's accurate AI data labeling

How Sama’s Accurate AI is Helping Blind Runners Run Independently

Google, through its early-stage research initiative Project Guideline, faced the challenge of developing technology that would enable people who are blind or have low vision to run independently without a human guide. The project required a highly accurate machine learning model capable of identifying a painted guideline under diverse and challenging real-world conditions. To build this model, Google needed a vast and varied dataset of precisely annotated images, a task they partnered with Sama to accomplish using its data annotation services.

Sama provided expert annotators who delivered the high-quality training data required. These annotators drew precise polygons around guidelines in images, learning to handle complex variables like obstructions and inferring line positions. This enabled Google to successfully train its algorithm, which was a key factor in the project's success. The solution allowed a blind CEO and marathon runner to run independently for the first time in over two decades, creating a life-changing navigation experience. A Google researcher credited Sama as a force multiplier, highlighting its high-quality, on-time delivery and flexibility.

Google

Xuan Yang

Computer Vision Researcher