The challenge
Our client, an American electric vehicle manufacturer and automotive technology company, required semantic segmentation for 2,000 images to support the development of their autonomous vehicles. Semantic segmentation helps specify the shape, size and form of the objects in an image, as well as their location. Looking to ramp the project with speed, the client turned to TELUS Digital to source the right talent.
The TELUS Digital solution
To fulfill our client’s requirements, we employed 22 young women from Noida, India who had sought employment through the SETU Foundation, an independent, voluntary organization that focuses on women’s education, youth empowerment and livelihood creation. Working with SETU, we established a detailed program to train the women on semantic segmentation. The annotators were tracked closely for precision and recall through daily quality checks and mentorship touchpoints.
The results
By the final weeks of the project, the team was exceeding quality thresholds and delivering ahead of schedule. Additionally, the group’s recall and precision levels exceeded benchmarks. Further, images containing nighttime scenarios and challenging road conditions, which can often make the annotator’s job more difficult, did not reduce the level of quality delivered.