The camera, designed for industrial and security applications, performs real-time AI processing directly on the device.
CamThink has launched the NeoEyes NE503 camera that features edge AI capabilities and operates entirely offline, performing computer vision tasks on the device itself without using any cloud computing or server services. The camera boasts up to 20 TOPS of AI processing power and is meant to be used in industrial automation, security and machine vision.
NeoEyes NE503 camera is equipped with the Hailo-15H processor which is a combination of a quad-core Arm Cortex-A53 CPU with dedicated AI acceleration. It has 8 GB of LPDDR4 memory and 64 GB of eMMC storage. It allows running several vision models offline while at the same time capturing 4K video. The platform supports OCI-compatible containerised applications on an embedded Linux environment based on the Yocto Project with the Python SDK, REST API, SSH access and support for Python, Go, and C++ development.
The camera incorporates the Sony IMX678 1/1.8-inch CMOS sensor that enables video capture at resolutions of 4K video at up to 60 fps. The other capabilities include HDR support, AI-based image signal processing, autofocus, and a motorised lens with an 8-32 mm focal range of 4x optical zoom.
The camera is capable of supporting computer vision applications like object detection, optical character recognition, face detection and recognition, person re-identification, pose estimation, behavior analysis, and attribute recognition. The camera comes preloaded with pre-trained models such as YOLOv8n used for person detection.
The camera can operate at temperatures ranging from -40°C to 60°C. It also comes with existing industrial and security systems including RTSP video streaming alongside structured event output through REST APIs and an internal event bus. Physical interfaces such as RS-485, alarm inputs and outputs, and Power of Ethernet for connectivity to access control systems, warning devices, sensors, and other industrial equipment.
Click here for the original announcement.
