IN Brief:
- WiseGuard combines an image sensor, neural processor, human detection model, and compression engine within a 13 × 11 mm module.
- Himax reports up to 98% human detection accuracy and more than five years of battery operation under its stated conditions.
- Evaluation kits are available for pre-order through DigiKey, with module shipments expected to begin in mid-November 2026.
Himax Technologies has opened pre-orders for its WiseGuard smart sensor evaluation kit through DigiKey, providing an engineering platform for low-power visual monitoring systems that perform human detection and object classification directly within the sensing device.
The WiseGuard module measures 13 × 11 mm and integrates an image sensor, neural processing unit, trained detection model, and image compression hardware. It is designed for security equipment, occupancy monitoring, connected buildings, and other applications that require continuous observation without transmitting every captured image to an external service.
The evaluation programme began on 3 November 2026, with shipments of WiseGuard module orders expected to start in mid-November. Himax describes the evaluation kits and module samples as a route for manufacturers to begin integration work before adopting the technology in finished products.
WiseGuard is based on the company’s established WiseEye low-power vision architecture. This combines image acquisition with dedicated processing hardware capable of performing selected machine-learning operations while limiting the electrical power required for continuous sensing.
Conventional camera systems can consume substantial energy when image acquisition, digital processing, and network transmission operate continuously. These demands become difficult to accommodate in battery-powered installations, particularly where equipment must remain active for several years without maintenance.
Moving detection and classification into the sensor module reduces the amount of information that must pass through the wider system. Instead of continuously transferring video to a remote processor, WiseGuard can analyse captured images locally and generate information describing relevant events.
The integrated neural processor performs the supported inference operations, while the trained model identifies humans and other objects. The resulting classifications can be supplied to surrounding application electronics, allowing the completed system to determine whether further processing or communication is required.
Himax reports human detection accuracy reaching 98%. This is a supplier-reported figure, and the company has not published a complete independent benchmark describing performance across every intended operating condition.
Detection accuracy depends on factors including the image sensor’s field of view, distance to the subject, ambient illumination, object position, and the characteristics of the scene. The frequency of false detections and missed events must also be considered when selecting a sensor for a particular installation.
A monitoring device used in a controlled indoor environment may encounter different conditions from a system installed outdoors, where changing light, weather, background movement, and partial obstruction can influence image quality.
WiseGuard’s local classification capability is intended to distinguish relevant activity from the broader stream of visual information. This can reduce unnecessary communication events, although the effectiveness of the approach depends on the selected model and how the completed system responds to its output.
The module also incorporates image compression hardware. Rather than transferring every captured frame at its original resolution, the system can select relevant visual information and compress it before transmission.
Himax describes a capability to crop and compress images associated with detected events. This allows a monitoring application to request visual confirmation without necessarily transmitting the complete field of view.
Cropping reduces the quantity of image data by excluding areas outside the selected region. Compression further reduces the amount of information needed to represent the remaining image, with the achievable reduction depending on scene content, image quality settings, and the compression implementation.
These operations introduce trade-offs. Excessive compression can reduce detail, while an incorrectly selected crop may exclude information needed for remote interpretation. The required image quality must therefore be established against the monitoring task rather than optimised solely for minimum data size.
The architecture also produces contextual metadata describing detected events. An application processor or connected service can use that information to determine whether to generate an alert, store a record, activate additional sensors, or request further information.
Such decisions are handled by the surrounding application system. WiseGuard supplies sensing and classification functions, rather than providing a complete autonomous control architecture for every connected device.
Low average power consumption is central to the product’s intended use. Himax states that WiseGuard can provide more than five years of operation from a single CR123 battery under its reference conditions.
The result depends on the operating duty cycle, frequency of detected events, temperature, communications activity, and battery characteristics. The company has not established that every finished product incorporating WiseGuard will achieve the same operating life.
Wireless transmission can be a substantial contributor to the total energy demand of a connected sensor. Reducing unnecessary image transfers can therefore complement low-power acquisition and inference, although the energy savings depend on the external radio and the amount of information transmitted.
Security functions are also integrated into the architecture. Himax identifies Arm TrustZone isolation, physically unclonable function-based authentication, image masking, and encryption among the available mechanisms.
TrustZone provides hardware-supported separation between selected execution environments, while physically unclonable functions can be used to derive device-specific security information. Their effectiveness depends on how the associated software, keys, and provisioning procedures are implemented.
Processing images locally can reduce the quantity of sensitive visual information transferred across a network. However, any images or metadata that leave the device must still be protected through appropriate communications, storage, access control, and retention mechanisms.
Samuel Kuo, Vice President of Smart Sensing Business at Himax, said the availability of evaluation hardware would support customer testing and integration, with the company targeting remote monitoring installations requiring infrequent maintenance.
Himax also identifies commercial surveillance, smart buildings, and occupancy sensing among the application areas under evaluation. The company has not disclosed named production customers, shipment volumes, or complete-system qualification results associated with the WiseGuard launch.
The DigiKey pre-order programme gives developers access to the evaluation platform, while the expected November module shipments provide a route towards hardware integration. Assessment of the detection model, average power consumption, communications overhead, and security configuration will determine the technology’s suitability for individual monitoring products.
