Boosting AI applications: Labeling format for complex datasets

摘要

Data labeling has become a major problem in industries aiming to create and use ground truth labels from massive multi-sensor archives to feed into Artificial Intelligence (AI) applications. Annotation of multi-sensor set-ups with multiple cameras and LIDAR is now particularly relevant for the automotive industry aiming to build Autonomous Driving (AD) functions. In this paper, we present the Video Content Description (VCD), as the first open source metadata structure and set of tools, able to structure annotations for such complex scenes, including unprecedented flexibility to label 2D and 3D objects, pixel-wise labels, actions, events, contexts, semantic relations, odometry, and calibration. Several example cases are reported to demonstrate the flexibility of the VCD.