M2I provides synchronized multi-camera observations together with 3D object annotations for infrastructure perception.
The dataset includes
Multi-camera RGB images
Camera calibration (intrinsics and extrinsics)
3D bounding box annotations
BEV semantic maps
Temporal sequences
Camera visibility information
Diverse road layouts (intersections, highways, roundabouts… See the full description on the dataset page:
https://huggingface.co/datasets/handsomeYun/M2I.