Advanced Perception Components v1.0.0

[bpapaspyros]

This is the first release of the Advanced Perception package.

It introduces the YOLO Executor component, enabling real-time object detection using YOLO models. For the first time, this package also depends on the AICA toolkits for machine learning and optional GPU acceleration.

Available Toolkits

• toolkits/ml
• toolkits/cuda

The ML toolkit images are open source and available at github.com/aica-technology/docker-images. We have also open-sourced helper scripts that make it easy to customize these images and use them natively with AICA Launcher v1.4.1 or higher. The new toolkits can be found under the Add package menu, similarly to other components and collections.

The toolkits/ml images include a broad range of Python and C++ ML packages, available in three variants:

1. CPU-only
2. GPU-first + CPU
3. GPU-first + CPU for NVIDIA Jetson

When using a GPU-compatible variant, you will also need toolkits/cuda, which provides the required CUDA (NVIDIA-only) libraries. If you forget to include it in a GPU configuration, the AICA package builder will remind you automatically.

About the YOLO Executor

The new components/advanced-perception package, and specifically the YoloExecutor component, supports both CPU and GPU variants. The CPU version is slower, but fully functional. Future versions may introduce components that require a GPU.

We are actively expanding documentation for all components and controllers. Stay tuned, a YOLO example is coming soon!

Dependencies

• package-builder: v1.4.0
• ros2-ws: v2.0.6-jazzy
• control-libraries: v9.2.0
• modulo: v5.2.3
• libpkgcheck: v4.0.0
• toolkits/ml: v1.0.0
• toolkits/cuda: v1.0.0 (optional)