Core Technologies
Multi-Source Data
Relying on the 4D data governance platform and the automated data production line, a closed-loop system covering real scene data, high-fidelity synthetic data and test vehicle data is constructed. Through the full-chain automated processing of data collection, cleaning, annotation, and simulation, high-quality training data is efficiently accumulated, accelerating algorithm iteration and model optimization, and providing continuous evolutionary impetus for the technical foundation.
In the data-driven era, the production efficiency of high-quality data is of vital importance. Based on the pre-training results, our company's large model can pre-label key elements such as parking Spaces, obstacles, and ground elements in 4D data. With our 4D annotation tool, all annotation results can be directly imported for preview and made open, including those projected back onto the original image. Users can view, inspect, and adjust and refine the annotations through the interface, effectively enhancing the production efficiency of 4D data.
注工具.mp4)
