Ambarella launches camera processor for real-time facial recognition on edge devices
Computer vision semiconductor company Ambarella has introduced a new camera system-on-chip (SoC) which powers real-time facial recognition on edge devices. The CV25 combining advanced video processing, high-resolution video encoding, and CVflow computer vision in a design the company says runs on extremely low power.
The deep neural network (DNN) processing provided by the new SoC also enables intelligent home monitoring, professional surveillance, and aftermarket automotive deployments such as smart dash cameras and driver monitoring systems, according to the announcement.
“CV25 brings computer vision at the edge into the mainstream,” said Fermi Wang, President and CEO of Ambarella. “With this new SoC, we are sharply focused on reducing our customer’s overall system cost for delivery of significant computer vision performance, high-quality image processing and advanced cyber-security features at very low power. CV25-based cameras are capable of performing Artificial Intelligence (AI) at the edge, allowing features like facial recognition to happen in real-time on the device, rather than in the cloud.”
The company notes that the CV25 can be integrated with smart doorbell systems to automatically recognize familiar faces, flag unknown visitors, and alert homeowners of package delivery.
Video is encoded in both AVC and HEVC formats at low bitrates to limit cloud storage costs, and the CV25 features a high-performance Image Signal Processor (ISP) for effective imaging in low light conditions. It also uses High Dynamic Range (HDR) processing for image detail extraction in high-contrast conditions, and includes a suite of cybersecurity features to protect against hacking. Ambarella says the 10nm ultra-low power processing technology makes the chip suitable for wireless camera applications requiring long battery life and small form factors.
Ambarella provides tools for customers to port their own neural networks onto CV25, including a compiler, debugger, and support for industry-standard machine learning frameworks like Caffe and TensorFlow, along with guidelines for convolutional neural networks (CNNs).