Sensor fusion is the integration of data from multiple sensor modalities to produce a more complete, accurate, and robust view of the environment than any single sensor can provide.
In autonomous driving and robotics, fusing 3D LiDAR with 2D cameras is common because LiDAR provides precise 3D geometry while cameras provide rich color, texture, and semantic cues.
For data annotation, fusion datasets require accurate sensor calibration and temporal synchronization so that extrinsic and intrinsic parameters align point clouds with images in both space and time.
Annotators often label objects in 3D first and then project the annotations into the corresponding camera views to verify consistency of 2D boxes or segmentation masks. This cross-modal validation helps keep annotations consistent across sensors and across frames.
LiDAR + Camera sensor fusion data annotation
