Holoscan Demos

This repository contains examples of how to use the Holoscan SDK and CyberEther to build SDR applications.

Examples

Allen Telescope Array Spectrum Analyzer

Holoscan Advanced Networking and CyberEther effectively turn the Allen Telescope Array into a large SDR. Go to example.

Complex I/Q Digital Signal Processing

Hello world example on how to manipulate complex IQ streams within Holoscan and CyberEther. Go to example.

Neural FM Radio

Example with an ONNX model directly demodulating an FM radio station from an SDR IQ stream. Go to example.

cyberether-holoscan-ml-fm-demo.mp4

Build Development Image

1. Clone this repository

$ git clone https://github.com/luigifcruz/holoscan-demos
$ cd holoscan-demos
$ git submodule update --init --recursive

2. Build base container

$ docker build -t holoscan-base -f Dockerfile-base .

3. Build demo container

$ docker build -t holoscan-demo -f Dockerfile .

4. Run the demo container

$ nvidia_icd_json=$(find /usr/share /etc -path '*/vulkan/icd.d/nvidia_icd.json' -type f -print -quit 2>/dev/null | grep .) || (echo "nvidia_icd.json not found" >&2 && false)
$ sudo docker run -it --rm -u root \
    --net host \
    --privileged \
    --runtime=nvidia \
    --device /dev/snd \
    --cap-add CAP_SYS_PTRACE \
    --ipc=host \
    --ulimit memlock=-1 \
    --ulimit stack=67108864 \
    -v /dev/bus/usb:/dev/bus/usb \
    -v /tmp/.X11-unix:/tmp/.X11-unix \
    -v /mnt/huge:/mnt/huge \
    -v .:/workspace/demos \
    -v $nvidia_icd_json:$nvidia_icd_json:ro \
    -e NVIDIA_DRIVER_CAPABILITIES=graphics,video,compute,utility,display \
    -e DISPLAY=$DISPLAY \
    holoscan-demo

5. Compile the examples

This directory with the examples will be mounted at /workspace/demos. You can compile the examples with the following commands:

$ cd demos
$ CXX=g++-11 CC=gcc-11 meson -Dbuildtype=debugoptimized build
$ cd build
$ ninja

6. Fun!

Check each example README for further instructions.

Notes

The official Holoscan v0.6 docker image comes with Ubuntu 20.04 and CUDA 11.6. But BLADE and CyberEther require at least CUDA 11.7 to work correctly. To fix this, we apply a patch to the Dockerfile distributed with the official image to update the base TensorRT docker image from 22.03 to 22.12. This will allow us to use CUDA 11.8 while still using Ubuntu 20.04 as the base image. This is a temporary fix until the code is ported to Holoscan v1.0 in the near future.

Feedback

An easy way to install Holohub operators. I had to hack an installer for the Advanced Network operator.
Holoscan complex number support. I guess this one is coming with an updated libcudacxx.