Computer Vision News - May 2022

8 Computer Vision Tool By Marica Muffoletto (twitter) Dear affectionate and new readers, welcome to a second chapter on MediaPipe . If you haven’t read the first one and you want a general overview of this tool, please check the March edition of the magazine here . In the previous article, we introduced this versatile machine learning framework, which is built with C++ and deployed to several platforms. We looked at the basic functionalities and the integration with Python. When embarking in the world of MediaPipe, the writer realised it was worth splitting the topic in two issues, to dedicate an entire part to the integration of MediaPipe with Android . It’s important to mention that similar tools have been developed for iOS and Javascript . If interested in those, please check out the tool website or feel free to get in contact for more information! Installing MediaPipe Let’s start with the installation of MediaPipe on your local machine. You can follow the steps below if you have a Linux machine with Ubuntu distribution (or visit here for Mac/ Windows/Docker installation). 1. Install Bazelisk from this link . Move into the desired workspace directory and clone MediaPipe repository: cd $workspace _ dir git clone https://github.com/google/mediapipe.git 2. Automatically build OpenCV cd mediapipe sudo chmod u+x setup _ opencv.sh ./setup _ opencv.sh 3. Run Hello World! In C++ to check installation export GLOG _ logtostderr=1 bazel run --define MEDIAPIPE _ DISABLE _ GPU=1 mediapipe/examples/desktop/ hello _ world:hello _ world MEDIAPIPE (2/2)