Ikomia Use Case: Facial Landmark Detection using OpenCV
– 5 min read –
Facial Landmark Detection is an important task in many applications such as emotion recognition in marketing, drowsiness detection in automotive or automatic facial makeup.
Today, we will explain how to detect facial landmarks in your images or videos using the Ikomia software and OpenCV. For a good introduction on facial landmark detection using OpenCV, we invite you to read this post from LearnOpenCV (Satya Mallick):
Well, wouldn’t be cool to reproduce these results with no-code on your own images or videos? Of course, YES! This is exactly the purpose of Ikomia and this is how it works. First we need to detect faces in your images/videos. Then we need to detect facial landmarks for each face.
How can we easily detect faces? One could use a cascade classifier, it’s fast and lightweight, or one could also use a deep learning based approach.
We choose to use a deep learning method since OpenCV already provides some pre-trained model, see PyImageSearch (Adrian Rosebrock) for a more detailed explanation:
In Ikomia, you just login into the software
Then, you open the Ikomia HUB, search for “face” and download the plugin “Face Detector”
Right now, your plugin is installed in your process library, close the HUB and use your new face detector on your images/videos like this
Then you can use the STUDIO to play with your plugin’s parameters.
Facial Landmark Detection
Once we have a Face detector, we can now detect facial landmarks on all detected faces. In Ikomia, open the Ikomia HUB and download the “Facemark LBF” algorithm as described above. Then you just apply the algorithm after the “Face Detector” and that’s it!
If you want to play with your facial landmarks detector on your video or stream, it’s pretty the same, just look at this video:
All plugins developed by Ikomia are Open Source:
Enjoy your Ikomia experience!
How to install Ikomia on Win10 and Linux.
Want to know more? Follow us on Twitter, LinkedIn and Youtube.
Please visit our website at https://ikomia.com.