Building High-Quality 2D Datasets for Autonomous Vehicles
About the Client
Client is known for providing commendable training data sets that are quite accurate. They opulent data sets are acknowledged for producing the greatest possible machine learning results in the development of autonomous vehicles. With a wonderful vision of technological advancement, they are trying to create futuristic data annotation services for all the clients.
Challenges
- Creating a comprehensive annotation-based flow with flag-off rules for robotic automobiles to identify signages/direction/traffic to be incorporated in ML development of robotic automobiles
- Creating a custom project to meet the autonomous vehicle training needs from 2D bounding boxes, to 3D cuboids, to lines and splines, polygons to landmarks and more, they required image classification and annotation services so that autonomous vehicles can navigate their environment in both rural and urban areas.
- With higher volumes to be delivered in a short duration of time with 99% accuracy was a challenge.
- Estimated 170+ people were required to be hired and trained in a week’s time.
Solution
- FiveS assembled a team of annotators and conducted a specialized training in less than 72 hours for them, to make sure they understand all the possible scenarios, disruptions to identify right attributes, classes and segments.
- The project consisted of multiple stages. First, we annotated lanes, traffic signals, signs and directions in videos with the help of bounding box, polyline, polygon, line annotation tools followed with classification of attributes.
- A quality framework with Maker: Checker was put into place to ensure accuracy of annotations and to ensure that the quality of annotation for the images were of the highest quality meeting the standards set by the customers.
Result
- 2.1M Annotations processed, analyzed and delivered
- 170+ Full-time data annotators completed the tasks in 108 days.
- 99% Accuracy levels and 100% quality checks.
FiveS provides the best customer experience by using multichannel contact centers, data-driven insight, shared devices and digital transformation by process automation. We also determine, design, implement analysis and improve business to ensure you can streamline and enhance back-office operations. We ensure a better level of attention to every customer inquiry.