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

[@byxlettera1452]

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

[@nmb-u-]

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

[@diegovalentino2083]

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

[@pastuh]

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

[@imtayyabhayat]

@ 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)

[@ThangVUHONG-u6d]

thanks man, your update help me solve the problem

[@pastuh]

Yep, appeared new parameter in FaceMesh MediaPipe, this solved problem

[@kshema6175]

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

[@mflow502]

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

[@АндрейБабаш-ю9п]

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

[@username-dh4tq]

.fill .folder .file : .pop fill. .file fill. folder file. .fill fill. .track fill. .slam fill. .pack fill. .bank fill. .dunk k. .net k. .txt k. .dot k. .hot k. .rod k. .pix k. .dis fill. .pop fill. .box fill. .cap fill. .vex fill. .bug fill. .cell fill. .lite fill. .txt fill. .url fill. .alt fill. .life fill. .fox fill. .cell fill. .poll pop. .fill folder. .fill fill. .fill

[@penjualanwilayah1826]

thanks bro regards from Syria

[@Jean-Naimar]

All your tutorials are great! Thanks

[@mh47saiyyad21]

Extremely excellent video thankyou sir.

[@malikabidrafiq]

Hi Murtaza, your video is great contribution for learning Python with computer vision. i was struck on an issues, incompatible function arguments. The following argument types are supported: 1. (arg0: bool) -> mediapipe.python._framework_bindings.packet.Packet then i replace following line self.faceMesh = self.mpFaceMesh.FaceMesh(self.staticMod, self.maxFace,self.minDetectionCon,self.minTrackCon) With self.mpFaceMesh.FaceMesh(static_image_mode=self.staticMod, max_num_faces=self.maxFace, min_detection_confidence=self.minDetectionCon, min_tracking_confidence=self.minTrackCon) this may helps other or guide me if something different .

[@pastuh]

Yes, useful info

[@arjunjeeyan1501]

Mind blowing amazing such thing your supper cool ma'am

[@viswatejvarma7792]

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

[@mh47saiyyad21]

Great knowledgeable video thankyou sir.

[@jordanellapin5655]

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!

[@arunappu3790]

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

[@muhammadnajamulislam2823]

AOA Bro FACE_CONNECTIONS is Replaces by FACEMESH_CONTOURS

[@_NguyenPhamTriToan]

Thank you so much !

[@MrBLAA]

THANK YOU!!!

[@Near010]

Thanks Bro!!

[@pastuh]

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

[@mohamedaitmouali895]

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

[@temyraverdana6421]

Amazing video. Thanks

[@mhmdfudhail383]

Make a video about facial expressions detection

[@securebrowser1479]

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

[@swedishpsychopath8795]

I'm trying to find it as well. Did you find it? For some reason it looks like google is trying to hide it. I can't find it anywhere and it kind of is the most important piece of information to have if you are going to use this framework to do your own stuff.

[@dhcocos]

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

[@Gh0st_0723]

cv2.videocapture (0)

[@TheRealFrankWizza]

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

[@keshav2136]

That will be good

[@dudeimfake]

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

[@aleksandraszczepanska1815]

amazing tutorial! thank you!

[@AndrewStifora]

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

[@tanishqsinghanand5383]

Hey , if u have fond them pls share .

[@nestorbao2108]

Also looking for the sample videos. Do you have any update on this?

[@murtazaburhani4022]

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.

[@theprogrammer8315]

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

[@Mindblowingclipss]

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?

[@DJPapzin]

🎯 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

[@sakurazzz289]

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?

[@OZtwo]

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

[@zeshan2973]

That's just an excuse. Many people don't even have Jetson Nano. But they keep doing this stuff

[@OZtwo]

@@zeshan2973 Well yeah but still the only issue I'm having is I can't get mine to update to the latest version without needing a second system to flash it. Soooo..as I try to fix mine I'm watching this very COOL video here.

[@zeshan2973]

@@OZtwo best of luck :)

[@OZtwo]

@@zeshan2973 ty, finding linux sucks. :)

[@ducnguyen4973]

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

[@leoyin2098]

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

[@Akshit21]

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

[@imalex4543]

That would require a trained model probably, and I don't think mediapipe provides one. Of course you could do something like `if the lips are higher than some other position, then say that the person is happy` but that would not be too accurate.

[@Akshit21]

@@imalex4543 but I'm trying to say that if we can use this model with any trained model then?

[@imalex4543]

@@Akshit21 I don't quite understand what you mean. Look, I had done research on how to make an emotion/mood detector in python, and could not find a module or trained model for it. Essentially, what you need to do is get a lot of pictures of happy faces and "feed" them to a "machine" that you have programmed in order for it to understand what a happy face is. Then it can detect the happy mood from any face. Edit: that is called Machine Learning or Deep Learning.

[@marsrocket]

Just google emotion detection. There are a bunch of examples of how to do it with OpenCV and keras.

[@navneetnsit09]

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

[@dilberismail]

Great job bro 👌👍

[@miku6921]

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

[@ahmedgr6392]

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

[@abquix]

I did it on my jetson nano. I installed latest mediapipe and latest other libs that mediapipe wants, like a bazel ...

[@PKTTV848]

I am going to integrate this to my

[@actionkey8042]

Murtaza's Workshop - Robotics and AI aweSOMME)

[@Life_Design_Ai]

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

[@Cyb3r-Kun]

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 :)

[@adesojialu1051]

Thanks boss, u r really a boss

[@ganeshgshet7797]

Fan of u broii ❤️❤️

[@sigmaphysicsclasses3232]

Good luck sir

[@juniasarias901]

You are the Best !!!!!

[@pamarthikanakaraja9812]

Super tutorial bro

[@danielrotnemer2564]

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!

[@dipeshdebnathltd]

google.github.io/mediapipe/getting_started/android.html go thru the docs here..it is given

[@makelabsindia3092]

nice sir and thank u

[@SUNYBOI]

Amazing

[@dracleirbag5838]

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.

[@bharatbhaishah8406]

How do you get such amazing ideas 💡

[@patrickknows2296]

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

[@CabrioDriving]

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

[@wgalloPT]

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

[@andihaki]

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

[@deependixit4214]

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!!

[@ErdosainNueve]

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

[@imtayyabhayat]

@ 12:30 FACE_CONNECTIONS has been renamed to FACEMESH_CONTOURS

[@kushandakshitha1682]

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

[@AnimeShorts-AP]

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

[@trinishmario1584]

which python version you are using?

[@Mohak-Bajaj]

Sir can we use this to augment faces over landmark positions

[@ruseboy]

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

[@kumarshivendu6011]

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

[@kwoksir2869]

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)?

[@CharmFlex]

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

[@MubashirAR]

Maybe it's possible to use this data to animate characters in an animating software like blender?

[@ilikamitra3429]

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

[@pierluigibuongiorno9464]

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

[@soothemeditationmusic2329]

facing same issue

[@maslovvitaliy4618]

удали 2 последних аргумента в __init__ и при вызове функции убери их

[@vipparthivijayababu8620]

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

[@ziaurrehman2180]

Pycharm

[@keshav2136]

Can u make a tutorial on MediaPipe?

[@yolodiy4242]

What IDE are you using? thanks

[@yolodiy4242]

saw your other vids, it's pycharm :D

[@gastonchevalet7342]

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

[@ANAmatör-rs]

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

[@thelaughingwood]

Where can we get sample videos used in this video?

[@nestorbao2108]

Any update? Have you found the sample videos?

[@thelaughingwood]

@@nestorbao2108 nope, i used real-time stream from my laptop's camera

[@fantomperfide5170]

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

[@Fin_raim]

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

[@flamelf]

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

[@murtazasworkshop]

Yes I will add the code soon.

[@hanns-erikquessel9102]

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 ?

[@c.jacquelin6523]

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

[@muhammadafifhidayat2566]

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

[@ran_css]

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

[@Fnta_discovery]

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

[@94142456]

Bro please share pycharm configuration settings

[@ItsKruski]

How do we download this software?

[@sohagmiah9839]

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

[@whatthefunction9140]

What can you do with this once detected?

[@manikantabandla3923]

The concept doesn't deserve to be taught for 40 min It's pretty simple

[@arjyabasu1311]

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

[@manikantabandla3923]

Why don't you deploy your code in Github?

[@ZukriBadwolf]

coz he has his own website , thus he uses its own rather than anyone else.....fully self dependent

[@pierluigibuongiorno9464]

where i can found a videos?

[@abdulrehmanajmal7200]

can you please provide data set :D

[@Saman6633]

best code

[@bahodirtohirjonov9231]

salom aka telegramga kanal chiqaring iltimos

[@nguyenanhnguyen7658]

What is it good for really ?

[@flamelf]

I'm using a slower version of this for modeling of 3D characters, but it is also how the snapchat filters and other overlays work

[@mr.zero-zone]

How to get face landmark without object images visualization

[@maaitrayodas5630]

how will I download the videos?