Ji Yue 01 rolls out first major OTA upgrade, debuts OCC-based pure visual intelligent driving feature

Beijing (Gasgoo)- On January 15, Ji Yue, the automotive robot (Robocar) maker jointly backed by Baidu and Geely, rolled out the first major Over-The-Air (OTA) upgrade for its Ji Yue 01 model since deliveries began. 

Photo credit: Baidu Apollo

The highlight of this update is the early experience of the Occupancy Network (OCC) feature, jointly developed by Baidu Apollo and Ji Yue. The upcoming V1.3 system will further integrate this technology, forming a comprehensive intelligent driving system of "BEV+OCC+Transformer," significantly enhancing the pure visual intelligent driving solution's capabilities for recognizing irregular obstacles and scene generalization.

At the core of the "pure visual" solution, OCC employs a deep learning-based 3D reconstruction method. This technology dynamically reconstructs the real world in 3D space through pixel cubes, accurately replicating various irregular objects in the car's visual world, such as obstacles, roads, buildings, and even plants. This enhances the intelligent vehicle's understanding of the surrounding environment.

OCC fills the final piece of the puzzle in Ji Yue's "pure visual" intelligent solution, granting the Ji Yue 01 initial capabilities for universal visual obstacle detection. With continuous usage, data input, iterative improvements, and refinement on the production line, the Ji Yue 01's perception abilities will visibly strengthen. Baidu Apollo envisions solving all autonomous driving perception-related issues using large-scale pure visual models in the future.

Powered by Baidu Apollo's technology, the Ji Yue 01 achieves detailed obstacle characterization, reaching centimeter-level 3D accuracy that rivals or even surpasses LiDAR precision. As pure visual technology is applied in-car, intelligent driving moves away from hardware-intensive approaches, reducing the costs for advanced urban intelligent driving functions and accelerating the functions’ widespread adoption.