Establish a professional image tuning team and develop proprietary patented technologies, including algorithm reconstruction for 2A (Auto Exposure&Auto Balance) for AI, optimization of ISP Pipeline, and improvement of post ISP processing such as LTM.

Based on CNN modeling, focusing on facial analysis in various digital vision methods to extract object feature values and engineering application screening and comparison.

Based on low-end hardware SOC in the industry, implement Face Detection, Face ID, Multi Face Trace, Sexual Attb. Recognition, as well as object recognition and tracking algorithms.

The algorithm is based on multi platform NPU optimization to achieve localized computing.

Design low-cost and low-power hardware architecture for multiple industry application scenarios such as commercial, industrial, and consumer.

Develop heterogeneous parallel computing and multi-threaded scheduling technologies for consumer grade and industrial grade SoC ARM, DSP, and VPU.

Combining the characteristics of cloud services, we collaborate with third-party cloud platforms to launch a commercial, low-cost, and efficient cloud edge facial analysis and computing solution.

In response to the characteristics of asymmetric low uplink bandwidth and low priority of asymmetric IoT network cards in the 4G IoT network environment, we have independently developed a multi-level data caching and block management mechanism based on the Davinc architecture design

We have launched the high reliability data stream transmission network protocol RTT PLUS for high-speed data transmission between multiple local devices.

We have improved the breakpoint resume protocol mechanism for cloud data transmission. In adverse network conditions, critical data can still be maintained without loss and transmission stability. 

At the same time, we will deepen cooperation with 4G chip suppliers and work together to optimize the underlying signaling code logic of cellular networks, thereby improving transmission efficiency and reliability.