Appen Limited, the leading provider of high-quality training data for organizations that build effective AI systems at scale, today announced enhanced capabilities to ensure autonomous vehicle manufacturers have access to high-quality training data and can get the most value from their training data investment. High-quality training data is essential to ensuring autonomous vehicles operate safely
and as expected, and Appen, which works with 7 of the 10 largest global automotive companies and tier 1 suppliers, can deliver 99+% accuracy for highly complex multimodal AI projects.
“It isn’t enough for vehicles to perform well in simulated or good weather conditions in one type of topography,” said Wilson Pang, CTO of Appen. “They must perform flawlessly in all weather conditions in every imaginable road
scenario they will encounter in real-world deployments. This means that teams working on the Machine Learning (ML) model for the vehicle’s AI must focus on getting training data with the highest possible accuracy before being able to deploy on the road. Our customers trust us with their most complex training data annotation scenarios because our industry-leading annotation platform and services enable us to deliver the high quality necessary to power multi-modal self-driving car algorithms.”
To understand and respond properly to road, weather, and safety conditions, autonomous vehicles require complex, multidimensional datasets from numerous and multiple types of sensors. This not only poses a challenge due to vendor specialization but also creates an enormous quality assurance challenge for the data annotation process because when teams that train the models to receive poor-quality training data, they must waste significant time and resources performing in-house audits to determine what parts of the datasets need improvement to provide a net benefit for their machine learning models. With more than 15 years of automotive industry experience, Appen’s data annotation teams regularly work with autonomous vehicle manufacturers to audit their existing annotated data and help them get closer to 100% quality, so they can get the most value from their training data.
For example, to enable their multi-modal autonomous vehicle ML algorithms, some manufacturers need to bind two distinct datasets of varying dimensions. This is extremely difficult to do manually but critical for autonomous vehicle model development. With Appen’s cutting-edge technology platform that delivers 3D point cloud annotation with object tracking by 99+% at cuboid level, customers can now annotate a dataset with 2D images bound to one with 3D point cloud annotations for mapping across multiple dimensions while aligning a consistent object ID requirement across 50+ frames.
“Our project is still in the pilot phase, and we needed to speed up the cycle to reach production, which requires training data that rapidly meets our algorithm requirements. The annotation tool, including 3D LiDAR, high-quality control features, and workflows, is already built into the Appen platform. This is helping us ensure the process is optimized based on our project requirements, enabling a smooth collaboration between our team and the Appen team. We are looking forward to moving this internal pilot into production,” said a senior project leader at Ecarx, an automotive technology company building an intelligent, connected platform for multiple vehicle models. The Appen training data platform combines human intelligence from over 1 million people around the world with cutting-edge models to create the highest-quality training data for ML projects. Appen is also committed to helping its customers ensure responsible AI – from pilot to production – based on ethical practices and data diversity, across all major use cases.