The new industrial edge AI camera combines Sony 4K imaging, 20 TOPS on-device inference, containerized application deployment, and system integration capabilities in a single device.
CamThink today announced NeoEyes NE503, a new edge AI camera platform designed for developers, system integrators, algorithm providers, and OEM teams building custom vision applications.
NE503 brings high-resolution image capture, local AI inference, application execution, structured event handling, and field integration into one industrial camera. The platform is designed for projects where a camera must do more than stream and record video: it must understand the scene, run project-specific logic, and connect results to software or physical systems.
Based on the Hailo-15H system-on-chip, NE503 provides up to 20 TOPS of INT8 AI performance. Its imaging system uses a Sony IMX678 1/1.8-inch CMOS sensor with 4K UHD output, HDR support, a motorized zoom lens, and a Gen2 AI-ISP for low-light and recognition-oriented imaging.
NE503 brings image capture, AI inference, application logic, and event output together inside the camera. It provides H.264/H.265 video for monitoring and evidence while exposing inference results to other services, enabling direct participation in security, industrial, parking, access, logistics, and AIoT workflows.
NE503 uses a containerized application architecture that gives developers control over models, application services, post-processing logic, and event handlers. Workloads can be packaged as OCI-compatible images and upgraded independently from the base firmware, helping algorithm vendors and OEM teams maintain their own software lifecycle.
The on-device AI runtime manages model loading, inference scheduling, and concurrent workloads. Supported vision workflows include object detection, OCR, face detection and recognition, ReID, pose estimation, behavior analysis, and attribute recognition. The platform also includes pre-installed models for rapid development and evaluation.
NE503 is designed to operate within existing video, enterprise, and industrial environments. RTSP streaming supports conventional video workflows, while the Event Bus and RESTful interfaces expose device and inference capabilities to applications. Alarm inputs, alarm outputs, and RS-485 enable local linkage with sensors, access systems, warning devices, and industrial equipment.
The software stack includes MediaClient, InferenceClient, and DeviceClient interfaces, together with a modular hardware abstraction layer. This separation allows application teams to work with media, inference, and device functions without tightly coupling business logic to low-level camera control.
NE503 supports a broad range of applications that combine professional imaging, local compute, and custom integration:
NeoEyes NE503 is now available for purchase. Initial deliveries are expected to begin in mid-July 2026 and will be fulfilled progressively. Demo units are available for product demonstrations, and teams can contact CamThink to discuss evaluation needs with our engineers.
CamThink welcomes inquiries from system integrators, algorithm companies, and OEM customers developing custom vision applications. CamThink engineers can assist with discussions around model deployment, imaging and lens requirements, containerized applications, system interfaces, and installation conditions.
Developers can explore NE503 specifications, system architecture, setup instructions, and software guides in the CamThink Wiki.
CamThink provides open and reliable edge vision AI hardware for faster prototyping, integration, and real-world deployment. Its NeoEyes camera platforms, NeoEdge edge AI computing products, and NeoMind edge AI platform help developers, system integrators, and OEM teams build visual AI systems with open software, practical interfaces, and deployment-ready hardware.
CamThink is backed by Milesight, a global IoT hardware manufacturer based in Xiamen, China. Learn more at www.camthink.ai.
Discuss your model, imaging, integration, installation, or custom application requirements with the CamThink team.
