Skip to main content

Developer Guide

Welcome to the Silight AI Camera Developer Guide. This section covers everything you need to develop applications for the Silight AI Camera AI Vision Processor.

Development Workflow​

flowchart LR
    A[Setup] --> B[Develop]
    B --> C[Test]
    C --> D[Deploy]
    D --> E[Optimize]

Core Components​

Video Pipeline (GStreamer)​

The video processing pipeline uses GStreamer with Hailo-specific plugins:

Camera -> ISP -> DSP -> NN -> Encoder -> Network

AI Inference (Hailo NN Core)​

Run neural network models on the dedicated AI accelerator:

ModelTaskPerformance
YOLOv5Object Detection100+ FPS @ 1080p
ResNet-50Classification1000+ FPS
MobileNetEdge AI2000+ FPS

Image Processing (ISP)​

Control camera settings via v4l2-ctl:

  • Auto Exposure
  • Auto White Balance
  • Auto Focus

Media Library API​

C/C++ API for building custom video pipelines:

#include <hailo_media_library.hpp>

// Create pipeline
PipelineConfig config;
config.load_from_json("/usr/bin/frontend_config.json");

// Start processing
MediaPipeline pipeline(config);
pipeline.start();

Prerequisites​

Development Host​

RequirementMinimumRecommended
OSUbuntu 20.04Ubuntu 22.04 LTS
RAM8 GB16 GB
Disk50 GB100 GB
Network100 Mbps1 Gbps

Cross-Compilation Toolchain​

# SDK location
prebuilt/sbc/sdk/poky-glibc-x86_64-*.sh

# Install
./poky-glibc-x86_64-*.sh -y

# Activate
. /opt/poky/4.0.23/environment-setup-armv8a-poky-linux

Hardware​

Setup​

Application Development​

  • YOLO Demo - Run object detection
  • WebUI - Browser-based camera control

SDK​

System Architecture​

Example Applications​

ApplicationDescriptionLocation
Object DetectionYOLOv5 demo/home/root/apps/detection/
Camera ViewerWebUI servercamera-viewer-server
Media LibraryAPI examplesGitHub

Development Tools​

ToolPurposeLocation
v4l2-ctlCamera control/usr/bin/v4l2-ctl
gst-launch-1.0GStreamer/usr/bin/gst-launch-1.0
camera-viewer-serverWebUI/usr/bin/camera-viewer-server

Debugging​

For development issues, see:

Next Steps​

  1. Review Technical Specifications
  2. Complete Cross-Compilation Setup
  3. Run the YOLO Demo
  4. Study the Media Library for custom development

Support​