City-scale vehicle-road-cloud fusion corpus: raw roadside multi-sensor streams (camera, LiDAR, radar) from 1,500+ instrumentised intersections, regulated-vehicle telemetry and trajectories, signal/phase data, high-precision map layers, plus curated perception, pre-training and abnormal-event/hazardous-condition scenario libraries in OpenSCENARIO-style formats
Equities in US, China/HK. Related tickers PONY, WRD, BIDU, NIO, MOFC, 9868.HK, 002594.SZ, 600104.SS.
Multimodal
More than 7 PB accumulated, growing about 450 TB per day (roadside roughly 6.3 PB, vehicle-side roughly 780 TB on one published split); curated sets include 500k+ images and 100k+ clips in the perception and pre-training libraries, 30,000+ road-surface data points, 10,000+ vehicle trajectories, and an earlier published 5.1 PB / 6 categories-200+ item inventory
Labels fall out of the operating mechanism: joint vehicle-side and roadside 2D/3D object annotations, per-intersection event tags (17 defined vehicle-road coordination event types, 8 abnormal classes such as automated-driving system exit, emergency braking, complex traffic flow), weather/lighting/road-surface attributes (7/5/3 categories), 10 road-obstacle categories, plate-and-operator identity, per-vehicle mileage and geofence compliance verdicts from third-party regulatory review, and safety-event adjudications
Beijing CarNet Technology Development Co., Ltd. (北京车网科技发展有限公司) — City-scale vehicle-road-cloud fusion corpus: raw roadside multi-sensor streams (camera, LiDAR, radar) from 1,500+ instrumentised intersections, regulated-vehicle telemetry and trajectories, signal/phase data, high-precision map layers, plus curated perception, pre-training and abnormal-event/hazardous-condition scenario libraries in OpenSCENARIO-style formats
Not yet on file: Sample · Data dictionary · Data frequency · Reporting lag · Point-in-time · License