Skip to main content

Cross-Compilation Setup

This guide covers setting up the cross-compilation environment and building applications using the Hailo Media Library (HML) for the Silight AI Camera.

SDK Download​

The software package contains the cross-compilation toolchain and board utilities. Download the Vision Processor Software Package from the resources section.

Package Contents​

hailo_vision_processor_sw_package/
├── prebuilt/
│   ├── sbc/
│   │   ├── sdk/
│   │   │   └── poky-glibc-x86_64-*-hailo15-sbc-toolchain-4.0.23.sh
│   │   └── ...
│   └── images/
│       └── hailo15-yocto-image.wic
└── tools/
    └── hailo15_board_tools_*.whl

Cross-Compilation Toolchain​

Install Toolchain​

Locate the toolchain installer in the SDK package:

# Navigate to SDK location
cd hailo_vision_processor_sw_package/prebuilt/sbc/sdk/

# Run installer (default path: /opt/poky/4.0.23)
./poky-glibc-x86_64-*-hailo15-sbc-toolchain-4.0.23.sh

Activate Environment​

Before compiling HML applications, activate the cross-compilation environment:

. /opt/poky/4.0.23/environment-setup-armv8a-poky-linux
tip

Add this to your shell profile or run it in each terminal session before building.

Hailo Media Library (HML)​

HML provides the API for developing video processing applications on the Silight AI Camera.

Download Source​

git clone https://github.com/hailo-ai/hailo-media-library
cd hailo-media-library

Compile HML​

With the cross-compilation environment activated:

meson build
ninja -C build

HML Application Development​

Configuration Files​

HML applications require JSON configuration files at runtime. You can use:

  • System presets: /usr/bin/*.json (on the target board)
  • Custom configs: Specify your own path when loading

Most JSON settings can be dynamically adjusted via HML API during program execution.

Frontend Config​

Example: /usr/bin/frontend_config_example.json

Configures video input and output processing:

SettingDescription
Vin/Vout resolutionInput/output frame dimensions
Frame rateFPS for each stream
DewarpGeometric distortion correction
Flip/MirrorImage orientation transforms
HDRHigh dynamic range processing
Low-light AI denoiseAI-powered noise reduction

Encoder Config​

Example: /usr/bin/frontend_encoders_sink*.json (sink0, sink1, ...)

Each encoder config corresponds to a Vout with matching resolution:

SettingDescription
Encoder formatH.264, HEVC (H.265), or MJPEG
Subtitle overlayText overlay on video
Image overlayLogo/watermark insertion
MaskPrivacy masking regions
Bitrate controlVBR (Variable Bitrate) or CVBR (Constant Variance Bitrate)

Loading Configuration​

#include <hailo_media_library.hpp>

// Load from system preset
PipelineConfig config;
config.load_from_json("/usr/bin/frontend_config.json");

// Or load custom configuration
config.load_from_json("/path/to/custom_config.json");

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

Dynamic Configuration​

Most settings can be adjusted at runtime via HML API:

// Example: Adjust bitrate dynamically
pipeline.set_encoder_bitrate(sink_id, new_bitrate);

// Example: Toggle HDR
pipeline.set_hdr_enabled(true);

Development Workflow​

  1. Install toolchain on development host
  2. Activate environment before each build session
  3. Develop application using HML API
  4. Cross-compile for ARM target
  5. Deploy to Silight AI Camera board
  6. Test with appropriate JSON configurations

Next Steps​