Facial Landmarks Detection | with Opencv, Mediapipe and Python

In this tutorial, we will see how to find 468 facial landmarks easily with a library called Mediapipe , extract the x and y coordinates so we can use them as we like.

➤ Full Video courses:

➤ Follow me on:

➤ For business inquiries:

#opencv #python #Mediapipe

[shinchan]

Thanks for this tutorial.
FYI; if any one getting the problem of loading the mediapipe library(DLL not found) ., then try to install msvc-runtime using pip.

[emo]

This information is exactly what I needed. Been playing around with the facemesh but couldn't figure out how to get the coordinate x, y to print out. Thanks you! liked and subscribed!

[Shinika]

Wow, really powerful library, thanks for introducing it to us ;)
Don't hesitate to provide more! totally worth it

[DJPapzin]

🎯 Key Takeaways for quick navigation:

00:00 🚀 *Introduction to Facial Landmarks Detection*
- Overview of the tutorial on finding 468 facial landmarks using the MediaPipe library.
- Sergio emphasizes simplifying the process for extracting facial landmark coordinates.
- Mention of the importance of having OpenCV and MediaPipe installed for the tutorial.
01:19 📚 *Loading and Displaying an Image*
- Instructions for loading an image using OpenCV and displaying it.
- Importance of verifying correct image loading by showing it.
- Sergio provides the necessary Python commands for image loading and display.
04:35 🛠️ *Initializing and Processing with MediaPipe*
- Explanation of initializing face mesh from MediaPipe for facial landmarks.
- Creating a face mesh object for detection and processing.
- Processing the image using MediaPipe's face mesh and addressing the RGB format difference with OpenCV.
07:42 🔄 *Converting Coordinates and Drawing Landmarks*
- Discussing the RGB format conversion for image processing compatibility.
- Extracting and converting x, y coordinates of facial landmarks for accurate positioning.
- Demonstrating how to draw circles on the image at each landmark point.
17:19 🎨 *Handling Multiple Facial Landmarks*
- Implementing a loop to access and draw circles for all 468 facial landmarks.
- Suggesting the use of text labels for better identification of landmark indices.
- Providing insights on making specific landmark points more visible.
19:26 📹 *Extending to Real-Time Video Processing*
- Transitioning from image processing to real-time video processing.
- Setting up a video capture object for webcam usage.
- Explaining the need for a while loop to continuously process frames in real-time.
22:24 ⚙️ *Handling Video End and Conclusion*
- Addressing potential errors when a video ends, causing an empty frame.
- Introducing an if statement to break the loop when there are no more frames.
- Sergio concludes the tutorial, expressing enthusiasm for further exploring MediaPipe functionalities in upcoming videos.

Made with HARPA AI

[suhani]

Thank god I found your video! It's so helpful for my project.

[dhanyaram]

pls upload a video on reenactment(talking face generation)

[drdev_blog]

perfect, make more videos about face landmark anylising!!!

[ericmlevy]

Once you have facial landmarks, what is the process of drawing polygons / connecting splines to create shapes called?

For instance “draw a white solid in the area under the eyes and another to the left and right of the mouth.”

[rashidabbasi]

please make a video, Point cloud compression for autonomous driving, Lidar please

[JominJose]

Thanks so much, this makes everything really easy

[rubenseror]

Thanks for the tutorial!
Is there a mapping between facial landmarks and region of interest?

[artsytechie]

this is cool thank you but could you possibly point me in the direction of information on how to use the points to project facial motion onto a 3D face model? Thank You!

[filltech]

hi i like all your videos . can you make video for face swapping using mediapipe instead of dlib from previous videos

[rajasekark]

how to save the facial landmarks in FBX format to use in maya or blender

[heshan]

Very nice video. Can you make a video on facial recognition using the same method? Or another way?

[informationDesigner]

Is it possible to use the z-coordinate to find out the direction of the head in which direction it is rotating... and can we know the real z-coordinate as x, y explained the z-coordinate?

[jesusmtz]

I see that mediapipe struggles if the video is much longer. I'm trying with imutils. Is opencv a good choice?

[yaminasfaxi]

hi
Thanks for this tutorial, i have a Project to detect a hand and i need to use the coordinate Z so i have some difficulty, any help please

[musicalworld]

Hi Thank you so much for the tutorial! Is it possible to map the 468 landmarks to 68 landmarks that dlib, opencv use?

[rabiaaziza]

Hi, I'm using Mediapipe for facemesh, and I'm stuck at how to compare facial measurements from different frames? Do you have any idea on how to do this please? thanks!