Detect 468 Face Landmarks in Real-time | OpenCV Python | Computer Vision 

Well done as usual. I found it the best way to start with face mesh. Clear and accurate. Thanks

The best tutorial i have seen on this subject, not too long nor too technical, i didn't knew i could inspect a code source by ctrl clicking it, and you have great best practices! thank you :^D

Wow, this tutorial is absolutely awesome! I never thought I could ever learn to do something like this. Thanks!

All your tutorials are great! Thanks

Pretty cool, I am going to integrate this to my cpp application.

your videos are very good and have a wide message thank you

Hey! It's quite impressive! I noticed the framerate can be quite low when you print the positions. So if I wanted get a stream video from a GoPro to send this data for example via OSC to Unity, I think I won't be able to animate a face at a high frame rate. I'm wondering because I currently use the iPhone 11 to animate a face at 60fps it's super fast but you MediaPipe demo has much more information that could be useful for facial motion capture. We may want to have at least 2 separate videos to capture 2 actors in live from a live stream from a GoPro. Do you have an opinion about that please? Thanks for the video it was really good to understand the process!

Amazing video. Thanks

amazing tutorial! thank you!

Wow, precise lecture. How can we know which number represents which landmark?

Extremely excellent video thankyou sir.

Mind blowing amazing such thing your supper cool ma'am

Great knowledgeable video thankyou sir.

Great job bro 👌👍

Great tutorial as usual. It would be really helpful for users if you could share the code.

Hello sir you are doing great Try to show us how to remove noise and blur in an image

thank you for your sharing. will you plan to have course about how to run this on web ?

Thank you for your share!I have a problem,what points are about lips,and what point about eyes.Can i find a sequence map about these point corresponding parts?

Can we use this to compare faces like for attendance system?

Fan of u broii ❤️❤️

thanks bro regards from Syria

nice sir and thank u

You are the Best !!!!!

Super tutorial bro

Murtaza's Workshop - Robotics and AI aweSOMME)

Really nice video :) Looks simple :X at 20:16 better enumerate, because later will be needed

thank you for your best explication you are the best how can i know landmarks of each position in the face like eye or mouth

Make a video about facial expressions detection

Thanks boss, u r really a boss

Amazing

Cool lessons))) Excellent work, explanation is excellent even for russian ukrainian speakers)

Fan of u bruh.. I have a question to visualise neural networks that we have developed to show in 3d?

Wow nice! Do you have some sort of image with the layout of each point on the face and their index?

How do you get such amazing ideas 💡

Can you adjust settings to interpolate landmarks between frames, so there is no jitter of points of interest on face?

You should do a tutorial on installing mediapipe on the jetson nano.

Thank you for you amazing videos ! I also tried to install mediapipe on my jetson nano but no luck , I tried for 3 days but couldn't find a solution

Sir can we use this to augment faces over landmark positions

Marvelous job! If i wanted to change this to real time capture (from a camera for example) what should i change?

You really need a good cpu and gpu for this. My pc give 30 fps max of just drawing and 4 fps when printing the points lol.

Can you put some machine learning and see if you need that many spots so you can reduce processes and ram use. It depends if your using it for just detection or emotions or recognition.

@ 25:15 3rd param is now 4th param, and at 3rd place there is new param called `refine_landmakrs` with a boolean value, so the whole function of FaceMesh will become like this FaceMesh( static_image_mode=True, max_num_faces=2, refine_landmarks=True, min_detection_confidence=0.5)

Hi i have a question! How Can i Print the x, y, z coordinates of one exact Landmark of the face? For Exempel : i want to Print the Position of my eyebrows. How do i geh this?

Interesting if number 1 (nose) is the main "static" point.. If yes, it can be helpful to measure each face zones

I hope you can help us operate our start up company and create beautiful products

I am a big fan Murtaza and I love your work. Have ever tried to use Rust for computer vision using the openCV bindings ?

Good luck sir

You are so lucky that you can play with all this! I can't even get my stupid Jetson Nano up and running fully! :)

Oh this is what i waited for! Is it possible to make this kind of face detection for Android? thank you so much for your videos!

Amazing!! Can anyone please tell me how can I set max_num_faces to infinity. I want to get all the faces from video. Thank you!!

Thank you so much for your python opencv videos. Where can I get the sample videos for this project?

Well ...can I use the landmarks data in 3d design software such as blender,3ds max , maya...etc

Is there anyway of recording and exporting the face mesh data?

Sir plz make a video on How to use mediapipe library in raspberry pi

if anyone has issues and is using new version of mediapipe, change 'mpDraw.draw_landmarks(img, faceLms, mpFaceMesh.FACE_CONNECTIONS)' to 'mpDraw.draw_landmarks(img, faceLms, mpFaceMesh.FACEMESH_CONTOURS)' worked for me as FACE_CONNECTIONS has been removed from mp.solutions.face_mesh

@ 12:30 FACE_CONNECTIONS has been renamed to FACEMESH_CONTOURS

AOA Bro FACE_CONNECTIONS is Replaces by FACEMESH_CONTOURS

I am going to integrate this to my

Can u make a tutorial on MediaPipe?

I could use this to know if I'm biting? I mean, do you think it would be possible detecte bruxism? Because, i can have the face relax around the mouth, and when the chin get close to the nose, then... I'm biting. It would be possible or is not the sufficient accurately? Thanks

can we detect only lips? i mean, can we make custom model for detecting lips keypoint / landmarks? thank you

can you make mood detection by this project? if yes then please make one cause it will be more accurate with this than other videos on youtube using different method

Can we build a Facial Recognition application by using these land marks ?

hay, this is an amazing tutorial and I thank you for making it. but I want to ask if you could make a video on how to get blendshapes with mediapipe? I would really appreciate it if you did also I'm subscibing :)

Hi is there a way I can add a 3D face mask on the face mesh in order to make face filters

How to show the 468 landmarks only by the captured camera video (that means only black background will be shown along with the 468 green landmarks)?

Thank you sir. Can you teach me how to run this facemesh on GPU? Does it automatically detect the GPU or any setting needed?

which python version you are using?

Hey can we make the face recognition system with this ?? 🤔🤔🤔

How to collect detailed color data from each coordinate of the face?

QUESTION: Why is it necessary to convert to pixels? It is already giving you coordinates....im sorry if its a dumb question...

Bro please share pycharm configuration settings

Is it possible to use this model in C++ ?

Is there any practical project that can be made of this?

Can we add the number of the nodes more than 468?

best code

can you please provide data set :D

which editor are u used to develop this plzz!! reply me

How can I detect the angle of rotation of faces ??

1. (arg0: bool) -> mediapipe.python._framework_bindings.packet.Packet how can i solve it

What IDE are you using? thanks

how will I download the videos?

🎯 Key Takeaways for quick navigation: 00:00 🚀 *Introduction to Face Landmark Detection* - Introduction to the video's topic: detecting 468 landmarks on faces using Google's model. - Mention of the premium course for creating real-world computer vision apps. - Setting up a PyCharm project and exploring videos for testing. 01:34 📹 *Exploring Videos for Testing* - Explanation of the videos folder containing various videos of different sizes. - Creation of a new Python file named "face_mesh_basics" for learning and testing. - Choosing a video and setting up the initial video capture. 02:18 ⚙️ *Installing Required Libraries* - Installation of OpenCV and MediaPipe libraries for face landmark detection. - Setting up the project interpreter in PyCharm. - Importing necessary libraries (cv2, mediapipe, time) in the Python file. 03:22 🎥 *Setting Up Video Capture and Display* - Initializing video capture using OpenCV (cv2) for the chosen video. - Creating a loop to continuously read frames and display the video. - Introduction to measuring and displaying frame rates. 04:43 📊 *Displaying Frame Rate on Video* - Measuring and displaying the frame rate using the time module. - Adding the frame rate information as text on the video. - Adjusting the font, color, and position of the frame rate display. 06:10 🤖 *Implementing Face Landmark Detection* - Initializing MediaPipe classes for drawing landmarks and detecting face mesh. - Converting BGR image to RGB format for compatibility. - Processing the frame to detect and draw face landmarks using MediaPipe. 11:17 🎨 *Customizing Landmark Drawing* - Adding flexibility by customizing landmark drawing specifications. - Adjusting circle radius and line thickness for better visibility. - Demonstrating the impact of customization on landmark visualization. 14:55 🔄 *Creating Face Mesh Module* - Transitioning code into a Python module for reusability. - Defining a class (FaceMeshDetector) with initialization and face mesh detection methods. - Incorporating parameters for customization and flexibility. 26:14 🔍 *Accessing Face Landmark Coordinates* - Exploring face landmark coordinates and their structure. - Converting normalized coordinates to pixel values using image dimensions. - Printing and visualizing the x, y coordinates of face landmarks. 27:44 🛠️ *Face Mesh Module Integration* - Integrating the face mesh module into the main code. - Utilizing the module to detect and draw face landmarks. - Returning the modified image for display or further processing. 28:15 🛠️ *Extracting Face Landmark Values* - Convert values to ensure proper return. - Drawing is optional, controlled by the "draw" parameter. 28:55 📝 *Processing Face Landmarks* - Loop through each face's landmarks. - Store x and y values in a list named "face." 29:50 🔄 *Handling Multiple Faces* - Create a list named "faces" to store landmarks of each face. - Append the "face" list to "faces" after processing each face. 30:38 🔄 *Returning the Result* - Return the "faces" list even if it's empty. - Ensure the code is outside the loop to avoid errors. 31:07 🖨️ *Printing Face Count* - Print the length of "faces" ifnot equal to zero. - Provides feedback on the number of faces detected. 32:17 🧐 *Verifying Landmark Points* - Check individual face landmarks using "faces[0]." - Useful for confirming if all 468 points are detected. 33:00 📊 *Displaying All Landmark Points* - Print the list of all 468 landmark points. - Useful for understanding the distribution of points on a face. 34:23 📌 *Adding ID Numbers to Landmarks* - Use cv2.putText to display ID numbers for each landmark. - Helps identify specific landmarks by their assigned ID. 36:28 📖 *Referring to MediaPipe Paper* - Suggests referencing the MediaPipe paper for detailed information. - Emphasizes the availability of information on point IDs in the paper. 36:56 🚀 *Achievement: Real-time CPU Processing* - Highlights the achievement of detecting 468 face landmarks in real-time on CPU. - Emphasizes the quality of results achieved. 37:42 🎥 *Testing on Different Videos* - Demonstrates smooth performance across various video samples. - Highlights adaptability to different scenarios, such as Zoom meetings. 38:43 🔄 *Flickering Issue Analysis* - Observes flickering, speculating on face merging as a possible cause. - Acknowledges potential limitations in certain scenarios. 39:11 🤓 *Conclusion and Call to Action* - Summarizes the video content and encourages learning. - Promotes engagement with a thumbs up and subscription. Made with HARPA AI

Hello, I could not run this project with rpi4. Could you help? (32 bit) ModuleNotFoundError: No module named 'mediapipe.python._framework_bindings'

I can't find the course on your website, is that still coming?

I want to install mediapipe on a Raspberry Pi 4 but it does not work : I always get the error; --> Could not find a version that satisfies the requirement mediapipe (from versions: ) No matching distribution found for mediapipe Can YOU help me? What is my mistake ?

How do we download this software?

What can you do with this once detected?

Where can we get sample videos used in this video?

How can I unlocked any projects by face detecting show us this project plz

where i can found a videos?

Can you please please make videos on how to create facial beautification(smooth skin, big eyes, skin tone) apps ? 😇

We here again haha

What is it good for really ?

Alhamdulillah

i'm confused because you repeate certain senteces twice example: drawSpec,drawSpec Why

salom aka telegramga kanal chiqaring iltimos

Pas de traduction française, dommage cela à l'air vraiment passionnant.

hello i want to make like u but with more Signs like if this human is happy or worried u can help me ?

How to open file in python IDLE?