NeoEyes NE503

More Than an IP Camera

A programmable edge AI camera platform for system integrators, AI solution builders, and OEMs to run vision models, application logic, and structured event workflows directly at the camera endpoint.

Get Evaluation Kit Video

Traditional IPC vs NE503

Traditional IPC Records. NE503 Turns Vision into Action.

Most IP cameras capture and stream video to servers. NE503 turns visual input into edge AI events. It runs models locally, applies application logic on the camera, and delivers structured results directly to your operational systems.

Highlights

Built Beyond Traditional IP Cameras

20 TOPS local AI compute powered by Hailo-15H

Structured event output through MQTT, Event Bus, REST API, and I/O

APP containers for AI models, business logic, and custom applications

Dual-board architecture separating AI compute and control

Open access platform with APIs, SDKs, apps, docs, and schematics

4K visual input with Sony IMX678, Gen2 AI-ISP, and motorized zoom

Wide operation temperature range from -30°C to +60°C

How NE503 Works

From Visual Input to Structured AI Events

Capture AI-Ready Visual Data

Structured events can be delivered through MQTT, REST API, VMS integration, or local I/O, making NE503 easier to integrate into existing monitoring and automation systems.

Sony IMX678 Gen2 AI-ISP Motorized Optical Zoom

NE503 capturing 4K visual data for edge AI analysis

Run AI Directly on the Camera

Powered by Hailo-15H, NE503 delivers 20 TOPS INT8 AI performance to run models locally at the camera endpoint, reducing server dependency and enabling faster on-site decisions.

Hailo-15H local inference

Local AI inference running on the NE503 camera

Deploy Custom Vision Apps

NE503 supports containerized applications, allowing integrators to package and deploy custom AI models, workflow logic, and connectors directly on the device for project-specific vision tasks.

OCI application isolated runtime independent updates

Containerized vision applications deployed on NE503

Turn AI Results into Events

NE503 filters model outputs, applies rules, and converts detections into structured events, so only meaningful outputs move forward to your systems.

event logic JSON payloads automation triggers

NE503 applying event logic to AI detections

Send Events to Your Existing Stack

Deliver structured events through MQTT, REST API, VMS integration, or local I/O, connecting camera-side intelligence to monitoring, automation, and field systems.

MQTT REST API RTSP Alarm I/O RS-485 Wiegand

NE503 connecting structured events to external systems

Visual Input Quality

AI-Ready Visual Input for Real-World Scenes

A Sony IMX678 4K sensor, Gen2 AI-ISP, and motorized optical zoom preserve the visual detail AI models need across low light, high-contrast scenes, and varying working distances.

4K full-color night vision scene with license plate and signage detail

4K Full-Color Detail at Night

Capture usable color detail in low-light scenes for detection, OCR, and scene understanding.

HDR balanced imaging in high-contrast backlight conditions

Stable Detail in Harsh Contrast

Keep foreground and background details visible when lighting conditions are uneven.

Motorized zoom range from wide scene view to target detail

Wide View to Target Detail

Adjust the field of view for both scene awareness and close-up visual recognition.

AI Compute

Multi-Model AI at the Camera Endpoint

NE503 orchestrates multiple concurrent AI workloads on the same video frame — Detection, OCR, Pose, ReID, Behavior, Visual Search — and outputs structured events for your system.

Multi-model vision processing pipeline with Hailo-15H chip, detection, OCR, pose, ReID, behavior, visual search, and structured event output

Apps Containerization

Deploy Your Own Vision Applications

Deploy project-specific AI models, workflow logic, and connectors as containerized apps directly on NE503.

OCI application architecture with AI model, business logic, system connector, and web UI on NE503

Package Your Workflow

Turn models, logic, and connectors into containerized applications.

Deploy as OCI Images

Install apps from a registry image or upload a local archive.

Run in Isolation

Sandboxed runtime keeps applications separated from system resources.

Upgrade Independently

Update models and applications without rebuilding the whole camera system.

See App Deployment Workflow

Supported Technology Stack

Build with familiar languages, package as OCI images, and bring models from common AI frameworks.

Python

C++

Golang

Docker

Linux

ONNX

PyTorch

TensorFlow

System Integration

Open Where Traditional AI Cameras Stay Closed

NE503 exposes video streams, structured AI events, APIs, and field interfaces so integrators can connect edge intelligence to VMS, IoT platforms, SCADA/BMS systems, backend applications, and on-site devices.

NE503 open integration diagram connecting VMS, IoT platforms, APIs, local AI apps, alarm I/O, and field devices

Connect Video Streams

Stream video to VMS, NVR, or your own vision pipeline.

RTSP Main Sub

Control Device Services

Manage device behavior through REST APIs and SDKs.

REST API SDK Storage

Route AI Events

Route AI results and device alerts into your systems.

JSON MQTT Alarm

Link Field Devices

Connect AI events to on-site systems and devices.

Relay RS-485 Wiegand

View Integration Guide

Hardware Foundation

Built for AI Workloads and Field Control

NE503 separates AI computing and field control into dedicated hardware layers, allowing vision applications to run, update, or restart on the camera while alarms, sensors, access control, and industrial interfaces remain reliably connected.

AI Apps run independently. Field control stays connected.

NE503 dual-board architecture with Core Board for Hailo-15H AI runtime and AI-PC Interface Board for alarm I/O, RS-485, and Wiegand

Deployment Hardware

Built for Real Fixed-Site Deployment

NE503 brings edge AI into real deployment environments with simplified PoE+ installation, IP66-class protection, passive cooling, wide-temperature operation, and field-ready I/O.

NE503 powered and connected through a single PoE+ cable

Single-Cable Deployment

Power and Data over PoE+

Use one Ethernet cable to power NE503 and connect it to the network, making fixed-site edge AI camera deployment faster.

PoE 802.3at Power + Network

NE503 operating outdoors in rain, dust, heat, and cold

Outdoor & Industrial

Built for Harsh Outdoor Conditions

A sealed enclosure and wide operating range support year-round deployment in exposed outdoor and demanding industrial environments.

IP66 -30°C to +60°C

Fanless NE503 thermal design for continuous edge AI operation

Continuous Operation

Fanless Cooling for 24/7 Edge AI

Passive thermal management removes a mechanical failure point and reduces maintenance for continuously operating fixed-site systems.

Fanless Design <8W Typical

NE503 connected to sensors, access control, alarms, and field equipment

Local Action

Connect AI Events to Field Devices

Trigger alarms, read sensors, and connect access-control or industrial equipment without adding a separate field gateway.

2× Alarm IN 2× Alarm OUT RS-485 Wiegand

Full Technical Specifications

Complete hardware specs, interfaces, and deployment parameters.

View Specifications

Edge AI Applications

One Camera Platform. Multiple Operational Workflows.

Use NE503 to detect, read, verify, and trigger actions directly at the camera endpoint—then connect results to the systems already running your site.

AI detection and local alerting for a restricted site zone

Security & Safety

Secure Critical Zones

Detect people or vehicles entering restricted areas, identify site-safety violations, and generate local alerts before full video needs to leave the site.

Perimeter Security Restricted Areas Site Safety

OCR and visual inspection of gauges, panels, and equipment states

Remote Inspection

Read Equipment States

Read gauges, panels, indicator lights, and equipment states, then convert OCR results or visual anomalies into timestamped operational data.

Utility Rooms Pump Stations Equipment Cabinets

License plate recognition and gate access event at a fixed entrance

Access & Traffic

Monitor Access and Traffic

Recognize vehicles and license plates at gates, verify entry conditions, and deliver access events to VMS, parking, or site-management systems.

Parking Entrances Campus Gates Logistics Yards