Establishing a More Intelligent World

Smart Vision Platform web

Smart Vision: A Clear, Insightful, and Responsive Way to See the World

To better serve global developers across a wide range of industries, HiSilicon offers high-performing chips and components, as well as core technologies such as 5G, AI, and 8K, helping construct versatile development platforms such as smart vision and smart IoT to address common requirements. HiSilicon provides developers with technical training and empowerment, communication platforms, and ecosystem cooperation services by engaging in independent design house (IDH) collaboration, developer community building, and initiating partnerships with universities, improving app development efficiency and nurturing innovation.

83% of the information obtained by humans through their senses comes from vision, and 11% comes from hearing, which explains why breakthroughs in artificial intelligence (AI) are most often related to image recognition. Machine vision has a broad range of applications, from manufacturing and security, to consumer electronics and transportation. More and more cameras now come equipped with AI image recognition capabilities. Visual data has become the most critical source of big data in the intelligent era. It accounts for over 80% of personal applications, and more than 70% of industry applications. HiSilicon has worked relentlessly on visual perception for over a decade, and accumulated industry-leading capabilities in such fields as Picture Quality (PQ) and AI. Our solutions offer professional-level image quality, while also setting the pace for PQ+AI, helping developers extract a wealth of value from video data, and explore groundbreaking product innovations.
Based on vision, AI, and connectivity, the HiSilicon smart vision development platform opens enriching visual processing, AI inference, and low power consumption capabilities, and integrates them into the HiSpark platform. By leveraging this platform, HiSilicon is able to provide optimal image quality, powerful computing capabilities, along with low power consumption, while making efficiency-boosting development tools broad accessible, for use in smart homes, consumer cameras, unmanned aerial vehicles (UAVs), and robotics, among other applications.

High image quality and low bit rate
  • Adaptive image enhancement, based on AI scenario identification, for improved imaging effects in complex lighting conditions, ensuring high-quality image output in challenging environments with low illumination and backlighting.
  • Smart video coding (SVC) technology, for lower operation costs, thanks to reduced backhaul bandwidth and storage usage

Efficient AI and effortless development
  • Heterogeneous high-performance AI and performance, power and area (PPA) computing engines, offering support for 1 TOPS computing power, to meet the wide-ranging computing power requirements in mainstream scenarios
  • End-to end (E2E) AI algorithm development and tuning tools, which simplify algorithm development and improve efficiency

Low power consumption and multiple ecosystems
  • A mature platform with a multitude of commercial applications, countless successful cases, and accumulated development experience
  • Industry-leading IPC+Wi-Fi low-power design
  • Multi-OS architecture (including Linux and Huawei LiteOS) and support for HarmonyOS, helping developers quickly import products that support the Harmony ecosystem
divider__line divider__line
poerceptual computing HiSpark web

HiSpark Development Kit

Visual processing solution:
The solution supports automatic face detection, and human profile tracking with a low false positive rate. It also features unique SVC policy that maintains image quality, while cutting the average bit rate for 1080p videos in half, reducing backhaul bandwidth and storage usage substantially. This solution provides visual perceptual capabilities for a wide range of different smart devices, serving as a home security monitor for children and the elderly in the home, or a dashcam in the car. For details about the development board, see

AI inference solution:
The solution combines visual perception with AI edge computing, and comes equipped with NPUs that offer 1 TOPS neural network computing power, quickly completing processes such as intelligent recognition and detection on the device side. Compared with traditional cloud processing solutions, it greatly reduces processing latency, with substantially lower security and privacy-related risks, since it does not need to upload visual data to the cloud. AI inference can be used in camera products that require strong AI computing power, such as AI cameras, smart cameras for large-screen devices, robotic vacuum cleaner, and facial recognition-based access controls. For details about the development board, see

Low-power solution:
This solution combines the vision module's turbo start function with the Wi-Fi module's keepalive mechanism. It features ultra-long standby time and provides visual perceptual capabilities for battery-powered smart home devices, such as smart doorbells, smart peepholes, and smart locks.

divider__line divider__line