Skip to main content

Documentation Guide

NE503 docs are organized around roles. Different roles need different content — this page helps you find your starting point: each role lists the core goal, the sections to read (links jump straight to them), and a suggested order. No need to read everything.

Documentation Sections Overview

SectionContentFor whom
Product InformationProduct positioning, core capabilities, interfaces and specsEveryone (a 5-minute product overview)
Quick StartLinear loop from unboxing to live (about 10 minutes)Evaluation / first deployment
User GuideAll web console features: image, peripherals, model management, ops, security, wiringOperations / installation staff
Hardware GuideBoard-level hardware: core board, interface board, development and debuggingHardware engineers
Software GuideSystem architecture, source build, flashing, version matrixPlatform developers
Application GuideApp development (SDK / tutorials / model training), Cookbook recipes, Reference (App / SDK / REST / video / events)App developers / integrators / model engineers
TroubleshootingSymptom-oriented troubleshooting FAQEveryone (look here when something breaks)

Pick a starting point by role

Evaluator / first deployment

Goal: from unboxing to a live device — verify camera, streams, and AI all work.

The path is the Quick Start (§1–§6, about 10 minutes). Once done, the device is live. For day-to-day use, see the User Guide.

App developer

Goal: build your own AI container app and deploy it to the device.

OrderSectionWhat it answers
1SDK prerequisitesSDK capabilities, how an app gets inference results
2First SDK callCode for subscribing to streams, inference, peripherals, events
3Declare permissionsDeclaring model and stream permissions in app.yaml
4Hello WorldBuild and run your first app end to end
5Cookbook: Parking LotWhat a complete business app looks like

System integrator

Goal: integrate the device's video streams, AI events, and management APIs into existing systems (NVR / VMS / platforms).

OrderSectionWhat it answers
1AuthenticationLogging in for a session token, using the integration key
2Video and ImagingRTSP URLs and parameters for NVR/VMS/FFmpeg
3Event integrationSubscribing to AI inference results and device alerts via WebSocket/MQTT
4REST API referenceAll management endpoints for models, apps, and the device

Model engineer

Goal: train your own detection model, compile it to HEF, and deploy it to the device NPU.

OrderSectionWhat it answers
1Model sources and choiceWhether preloaded models suffice or you need the full pipeline
2Version matrix · factory preload listWhat models the device ships with, component version constraints
3Model trainingYOLOv8 training: dataset, 384×640 input, hyperparameters
4Hailo HEF quantized compilationONNX → quantization → NMS → HEF end to end
5Deploy to NE503Import to device, verify loading, run end to end

Platform developer

Goal: understand and build the entire platform software stack.

OrderSectionWhat it answers
1Four-layer platform architectureHow platform services / HAL / apps are layered
2End-to-end data flowThe full path of a frame from sensor to event
3Developer guide · Environment setupThree-repo layout, toolchain, Docker environment
4Version buildBuilding platform components from source

When something goes wrong

Whatever your role, start troubleshooting in Troubleshooting: a symptom-oriented FAQ (device & network / video & streams / AI & models / apps & containers / events & integration / storage & disks / flashing & peripherals / system & services) organized as symptom → cause → fix.