No products in the cart.
NeoEyes NE503 4K PoE Edge AI Camera
📦 Ships from Late-July
NeoEyes NE503 is a 4K edge AI camera for advanced video analytics, smart security, industrial inspection, and AIoT systems.
Built on the Hailo-15H SoC, it combines 20 TOPS on-device AI inference, Sony IMX678 4K imaging, Hailo Gen2 AI-ISP, 8–32mm AF auto-zoom optics, RTSP streaming, structured event output, and containerized AI application deployment in one IP67 PoE camera.
More technical details are available in the NE503 official wiki.
Overview
CamThink NeoEyes NE503 is a 4K edge AI camera for local video intelligence, outdoor deployment, and custom AI application development. Built on the Hailo-15H SoC with a 20 TOPS INT8 NPU, it runs inference directly on the device and publishes structured events through Event Bus or MQTT. See the NE503 official wiki for the full technical overview.
NE503 combines Sony IMX678 4K imaging, Hailo Gen2 AI-ISP, and an 8–32mm F1.6 AF auto-zoom lens for high-quality monitoring, low-light imaging, and long-range recognition. For setup, RTSP streams, and application deployment, refer to the NE503 Quick Start Guide.
For developers and integrators, NE503 supports containerized applications, REST API, Python SDK, Web Console, SSH, and aipc-cli. Technical users can review the system architecture and SDK workflow before deployment.
Key Features
Hailo-15H 20 TOPS edge AI platform
The Hailo-15H SoC provides 20 TOPS INT8 AI performance with a quad-core Arm Cortex-A53 CPU and 350 GOPS DSP. NE503 can run visual AI workloads locally on the camera and output structured inference events without sending every decision to a cloud server.
On-device AI inference
NE503 runs AI workloads locally on the camera, reducing cloud dependency, network bandwidth usage, and response latency. The platform supports multi-model concurrent inference and outputs structured inference results through Event Bus for local linkage or system integration.
Supported AI task types include detection, OCR, face detection and recognition, ReID, pose estimation, behavior analysis, and attribute recognition. Pre-installed models include Person Detection YOLOv8n, Face Landmarks, and CLIP Image Encoder. See the NE503 AI inference overview.
4K imaging with AF zoom optics
NE503 uses a Sony IMX678 4K CMOS sensor with Hailo Gen2 AI-ISP. The 8-32mm F1.6 AF auto-zoom lens provides 4x optical zoom, with a 45.1° horizontal field of view at 8mm and 14.7° at 32mm. This makes it suitable for both wider scene coverage and longer distance recognition. Hardware details are available in the NE503 hardware specifications.
Field of view
| Mode | Horizontal | Diagonal | Vertical |
|---|---|---|---|
| Wide 8mm | 45.1° | 52.8° | 24.6° |
| Tele 32mm | 14.7° | 16.8° | 8.4° |
RTSP video for VMS and NVR integration
NE503 provides main, sub, and third RTSP streams. Use the main stream for 4K recording, the sub stream for lower bandwidth preview, and the third stream for AI analysis or mobile use. The default RTSP port is 8554. Learn more in the NE503 Quick Start RTSP section.
Containerized AI application platform
NE503 is not only a smart IPC. It is also an open edge computing platform. Developers and system integrators can deploy AI applications using containerized workflows, manage models and applications through the Web Console, and integrate with external systems through REST API, MQTT/Event Bus, SSH, Python SDK, and aipc-cli.
Application deployment and model management are covered in the NE503 Quick Start guide or NE503 Application Reference
For application packaging and SDK usage, follow Build AI Apps on NE503.
Model management and local inference output
NE503 ships with pre-installed models including Person Detection YOLOv8n, Face Landmarks, and CLIP Image Encoder. Custom Hailo HEF models can be imported through the Web Console or registered with aipc-cli. The NE503 Person Detection Demo is a useful starting point for teams building their first camera-side AI app.
Outdoor PoE deployment
NE503 supports PoE 802.3AT or DC 12V input, typical 5-6W power consumption, IP67 protection, RS-485, Alarm IN x 2, Alarm OUT x 2, TF card expansion, and operation from -40°C to +60°C. PoE is recommended when the installation needs power and network through one cable.
Where NE503 Fits
Smart security
Perimeter monitoring, person detection, license plate workflows, VMS integration, and local event output.
Industrial inspection
On-device visual inspection, safety equipment detection, alarm linkage, and custom model deployment.
AIoT applications
Camera-side inference, structured data output, REST API integration, and MQTT/Event Bus workflows.
Legacy IPC upgrade
Use NE503 when an existing video system needs local AI capability beyond standard RTSP video. For buyer education, see IP Camera vs Edge AI IPC Buying Guide.
What’s Included
- 1 × NE503 Camera
- 1 × Wall-mount
Customer-provided
- PoE 802.3AT switch and Ethernet cable, recommended
- Or DC 12V power adapter, standard network switch, and Ethernet cable
- Computer with Ethernet port and modern browser for first-time setup
- Optional TF card if additional local storage is required
Purchase Notes
- M.2 SSD hardware is documented, but it is not yet supported by the current firmware.
- RTSP is the current video integration protocol. RTMP is planned.
- Change the default Web Console and SSH credentials before deployment.
NeoEyes NE503 Specifications
Core Platform
| Product Model | NE503 |
| SoC | Hailo-15H Quad-core Arm Cortex-A53 @ 1.3 GHz 20 TOPS @ INT8, with 4-bit quantization support |
| DSP | 350 GOPS |
| Memory | 8 GB LPDDR4 |
| Storage | 64 GB eMMC, support TF card |
| Operating System | Embedded Linux, Yocto build |
Imaging System
| Image Sensor | Sony IMX678, 1/1.8" CMOS |
| Effective Resolution | 3840 × 2160, 4K UHD |
| HDR | Digital Overlap / Dual Gain HDR |
| Lens | AF auto-zoom lens |
| Focal Length | 8–32mm |
| Optical Zoom | 4x |
| Max Aperture | F1.6 |
| Field of View | H 44.5 deg (W) / D 52 deg (W) 14.5 deg (T) / 16.6 deg (T) |
| Lens Control | AF autofocus and auto-zoom |
| Image Stabilization | EIS supported by onboard gyroscope |
| Low-Light Imaging | <0.01 LUX full-color with AI-ISP |
Learn More in NeoEyes NE503 Core Board
Video & AI
| Video Encoding | H.264 / H.265 hardware encoding |
| Max Encoding | 4K@30fps |
| Rate Control | CBR / VBR |
| RTSP Streams | Main stream / Sub stream / Third stream |
| Inference Latency Target | <50 ms |
| Concurrent Inference | Multi-model simultaneous execution supported |
| Data Pipeline | Zero-copy DMA/SHM pipeline |
| Pre-installed Models | Person Detection YOLOv8n, Face Landmarks, CLIP Image Encoder |
| Supported Tasks | Detection, OCR, face detection/recognition, ReID, pose estimation, behavior analysis, attribute recognition |
| Event Output | Structured inference results, event messages, device status data |
Connectivity
| Ethernet | 10/100M LAN |
| PoE | IEEE 802.3AT |
| Video Protocol | RTSP |
| Data Protocols | MQTT / Event Bus |
| Planned Protocol | RTMP |
| External I/O | Alarm IN × 2, Alarm OUT × 2 |
| Serial Interface | RS-485 |
| Management | Web Console, SSH, REST API, aipc-cli |
| Developer Tools | Python SDK, REST API, aipc-cli, Hailo Dataflow Compiler workflow support |
Learn More in NeoEyes NE503 Interface Board
Deployment Environment
| Power Input | PoE 802.3AT or DC 12V |
| Typical Power Consumption | 5–6W |
| Protection Rating | IP67 & IK10 |
| Operating Temperature | -40°C to +60°C |
| Relative Humidity | 0–95% RH, non-condensing |
| Installation | Wall-mount screw pack included |
NeoEyes NE503 Interface Board Layout

NeoEyes NE503 Main Board Layout



























