Minibot Vision

This repository contains the sources to evaluate camera input for our minibot. The following workshops are now supported:

• Traffic sign workshop

Setup

This project comes together with the minibot project. Check out the readme of this project for further information.

Todos

• GPU support for opencv

Available docker images

Getting started

To run the core docker image with tensorflow, ROS noetic and minibot-vision with GPU support for tensorflow, execute the following command:

sudo docker run -it --runtime=nvidia -v /usr/local/cuda-10.2/targets/aarch64-linux/lib:/usr/local/cuda-10.2/targets/aarch64-linux/lib:ro --rm docker.uibk.ac.at:443/informatik/stair/minibot-vision/arm

This creates an interactive container starting with the shell as entrypoint. After the container has been stopped it will be deleted.

Building docker container

Base Image with Tensorflow and ROS noetic

To run the sources of this repository on our minibot with GPU support a specific docker image (nvcr.io/nvidia/l4t-tensorflow:r32.7.1-tf2.7-py3) is required. This image has been used to create an image with ROS noetic on Ubunut 18.04. The final image with tensorflow and ROS noetic is available on the UIBK docker registry: docker.uibk.ac.at:443/informatik/stair/minibot-vision/ros/arm

Build image from Dockerfile

Navigate to the directory .docker and execute the following command on the minibot:

sudo docker build --tag docker.uibk.ac.at:443/informatik/stair/minibot-vision/ros/arm .

If you want to build the image on another device (not arm64/v8) the following command is required:

sudo docker buildx build --platform=linux/arm64/v8 --tag docker.uibk.ac.at:443/informatik/stair/minibot-vision/ros/arm .

Core image with minibot vision

Based on the docker.uibk.ac.at:443/informatik/stair/minibot-vision/ros/arm image a docker image with cv_bridge and the minibot-vision sources complied has been created: docker.uibk.ac.at:443/informatik/stair/minibot-vision/arm. Inside the container the minibot-vision sources are available in the directory/app/catkin_ws/src.

Build image from Dockerfile

To build the image with the sources of the repository and using the base image the following command is required:

sudo docker build --tag docker.uibk.ac.at:443/informatik/stair/minibot-vision/arm .

If you want to build the image on another device (not arm64/v8) the following command is required:

sudo docker buildx build --platform=linux/arm64/v8 --tag docker.uibk.ac.at:443/informatik/stair/minibot-vision/arm .