HoloMatic's visual perception algorithm is deployed on Texas Instruments chips, one step closer to mass production

Recently, the autonomous driving startup Hodo Technology announced that its developed visual perception system has been deployed on an embedded platform based on the Texas Instruments (TI) TDA4VM chip. Judging from the deployment results, Hodo Technology's visual perception algorithm system is relatively mature, and the solution can meet automotive grade requirements, achieving support for the mainstream product line in the market, and taking another important step on the road to mass production of autonomous driving.

For autonomous driving systems, the combination of software and hardware is very important, which will directly affect the release of computing power and the use of algorithms, and ultimately affect the performance of the system. The deployment of software algorithms on automotive-grade chips has always been a difficult point in the autonomous driving industry.

As an important foundation of autonomous driving technology, the ability to perceive the environment is the focus of many companies. Hodo Technology adopts a multi-sensor fusion solution at the mass production level, which requires a powerful visual perception algorithm to more efficiently process the massive data collected by multiple sensors. Previously, the visual perception algorithm independently developed by Hodo Technology ranked first in the "monocular 3D obstacle detection" ranking on the authoritative autonomous driving dataset KITTI with higher accuracy and robustness.

Texas Instruments (TI)'s TDA4 series chips have been selected as computing platforms by many automobile manufacturers and Tier 1 suppliers. The TDA4VM developed based on Jacinto™7 is an automotive-grade chip launched for the next-generation intelligent driving applications. It focuses on the high-end autonomous driving market and has the advantages of strong performance, low cost, low power consumption, and high safety.

Although TDA4 has the aforementioned advantages, as a low-power automotive-grade chip, its computing power is always limited. This requires software algorithm developers to have strong control over the system, a deep understanding of the computing platform infrastructure, and excellent algorithm design and compression capabilities to successfully deploy complex autonomous driving software algorithms on this automotive-grade chip and achieve excellent autonomous driving performance.

It is understood that after the deployment of the TDA4VM chip, the road test results of HoloMatic's visual perception algorithm showed that the visual perception algorithm system ran stably and could achieve stable perception results under complex road and traffic conditions, realizing core functions such as monocular visual obstacle detection and tracking, and road feature detection, reflecting the characteristics of high detection accuracy, high real-time performance, and accurate tracking.

HoloMatic was founded in 2017. Based on cutting-edge artificial intelligence technology and automotive industry technology, it has built a complete layout from artificial intelligence algorithms to embedded systems, from big data closed loops to system iterative evolution. Currently, HoloMatic is working with a number of automakers to promote the mass production of HoloPilot, an autonomous driving system, and HoloParking, an intelligent parking system. In the future, HoloMatic will continue to strengthen its technology accumulation in perception, positioning, planning, control, etc., and plans to put autonomous driving systems into mass production of new models this year together with a number of partner automakers.