AI smart camera chip vision module
The AI ??chip vision module consists of Sensor+ISP+VPU+ embedded depth learning vision algorithm, which can output structured data from the chip end. The module measures only 38mm x 38mm. It is understood that the module can carry the algorithm IP for different scenarios, and currently mainly includes the star-face recognition module, the star-data acquisition module, the star-human interaction module.
The Star Module greatly reduces the barrier to integrated visual recognition. For example, through a simple integrated star-face recognition module, a normal camera can be turned into a smart security camera with real-time structured processing capabilities, with powerful face capture and recognition. Ability to track and capture 30 people at the same time, grab the best face with the clearest angle, capture rate 99%, false detection rate less than 0.5%, capture repetition rate less than 10%, and support full locals within 20,000 Face recognition supports local feature extraction and transmits the cloud to meet face recognition scenes of more than 20,000 people.
In addition, the Star Module is turning the camera into a super data sensor through its front-end processing capabilities that rival the cloud. As an example, the face-to-face technology developed a smart customer group analysis camera based on the Star-Data Acquisition module. Based on the star’s local data structuring capabilities, the reader can easily perform real-time passenger flow counting. Member/customer management track management and other functions, the merchant can grasp customer attributes and behaviors in real time.
From the perspective of human-computer interaction, the Star-Human interaction module can be flexibly used in intelligent machines. By detecting the gesture in the image or video stream, the human face is transformed into instructions that cause the machine to respond correctly. It achieves millisecond detection and tracking speed, and responds quickly. It supports palm punches, V-shaped hand, stretched palms, fists, and other gestures. It is highly robust for face and human detection tracking.