UPDATE ON VIDEO I understand that the GitHub of mediapipe-bin has changed and you guys were getting errors while installing mediapipe , so I have fixed the error and changed the last step in my Github. so now there should be no error and you can get mediapipe running in the jetson with no problem. Thank you
I have gradually completed the steps in the video, but I am still encountering the following message.The device I am using is Jetson Xavier NX with CUDA 10.2 and Ubuntu 18.04, but I am still encountering the message ‘created TensorFlow Lite XNNPack delegate for CPU,’ and the FPS remains between 8 and 9.”
Let me check on that as it should automatically run mediapipe on GPU as I made this video long back and tested on both jetson nano and jetson Xavier , so the updates on jetpack may caused changes in the installation.
@@priotx I appreciate your prompt response. When I reached Step 5 in the instructions provided at this link, I encountered an issue where the packages numpy-1.19.4-cp36-none-manylinux2014_aarch64.whl and mediapipe-0.8.5_cuda102-cp36-none-linux_aarch64.whl are no longer supported. Thank you for your assistance. My python version is python3.8
Post inatllation, how do I confirm if the installation is successful and also how do I confirm if a program is running on the GPU or CPU on the jetson nano ?
@@DavideOrlando1969 there are videos in RUclips about what you like to do, I will be making videos in my channel also and bring mediapipe tutorials soon. thank you
Hi, I followed the newer steps as mentioned in your GitHub. In step 5 after I install the mediapipe-bin.zip from google drive do I need to come out of directory “cd mediapipe” and then do “cd downloads” and unzip it there? Also in “cd mediapipe-bin” has to be run inside “cd downloads”?
Thanks all your videos! I love them pretty much! Currently, I've trained a model in Google COLAB, but the trained model generated is in PYTORCH. So, I'd like to know what I have to do to run this model on the Jetson NANO. Thank you in advance.
this may help you - pytorch.org/blog/running-pytorch-models-on-jetson-nano/ I will make a video on this topic, I am busy till May 1st. so I will the make video after May 1 st.
I was wondering if I can do this on a virtual environmental (python 3.8) on jetson nano, because using sudo to install things means that it is not installed on my virtual environment but is actually installed on my original environment
yes you can do it on virtual environmental , but the main dependency should we installed on both root and virtual environmental , but I recommend to run it the normal way
Does this video include building Mediapipe on the jetson nano with GPU support ? I'm trying to implement a facial detection+recognition project and I want the facial detection part of the project to run on the GPU using Mediapipe. After following the steps in the video above will I have Mediapipe built with GPU support?
@@priotx Ok, I'm currently running a facail detcetion+recognition project using OpenCv and DLib, it runs at about 5fps for 2 webcams. Post Mediapipe installation, where I will be running the detection on the GPU using MediaPipe while the recognition part of the project continues to run on the CPU, Will there be a improvement in the overall frame rate?
This is specially for jetson nano , but if you what to try mediapipe on windows it is really easy to install and run , check other videos on youtube you will find then
@@user-lo7ck7bh1i for linux or ubuntu there is a much easter method of installation on mediapipe check other videos you can find then online , after installing the same code will work on both linux and jetson nano perfectly
Hi, I followed the instructıon in github link with the bin file you shared. I reached the last command in the github link but after that I don't know which commands I have to do? from which command in the video I need to continue to the end?
Hi Melvin thanks for the explanation, I've been having trouble with the command sudo pip install opencv_contrib_python this never ends and stays in the bulding wheel for opencv-contrib-python(pyproject.toml) if you or anyone can help me, please.
@@codaemon375 I undersand that , the github of mediapipe-bin has changed , I will update it soon in my github also , I have the solution , but need to make sure it is working fine and I will update it in my GitHub. so please wait , I will fix it soon
@@codaemon375 I have upadated the github , can you try to install mediapipe now using the new way I have shown in the github and tell me it is working or not
UPDATE ON VIDEO
I understand that the GitHub of mediapipe-bin has changed and you guys were getting errors while installing mediapipe , so I have fixed the error and changed the last step in my Github.
so now there should be no error and you can get mediapipe running in the jetson with no problem.
Thank you
Thanks! Left a sub on your channel
from next month I will be putting more videos , on robotics and computer vision
Awesome Content, made it first try, mine took 8 HOURS at opencv_contrib_python command however, also used about 6.8GB total RAM
good to know that this video helped you
Melvin Machane adi pwoli video !!
thanks , how did you know I am from Kerala
@@priotx eda ninke enne manasilayile ?
@@immanueljzv6832 are you in Instagram or telegram lets talk there , just search Melvin Sajith in both you can find me there
Great to see it works also on Jetson. Curious to see what you do with it.
You can do all the project using mediapipe on the jetson it works very well
Awesome. Glad I stumbled into this video.
Glad you enjoyed it!
If this video helped you , please subscribe to my channel
I have gradually completed the steps in the video, but I am still encountering the following message.The device I am using is Jetson Xavier NX with CUDA 10.2 and Ubuntu 18.04, but I am still encountering the message ‘created TensorFlow Lite XNNPack delegate for CPU,’ and the FPS remains between 8 and 9.”
Let me check on that as it should automatically run mediapipe on GPU as I made this video long back and tested on both jetson nano and jetson Xavier , so the updates on jetpack may caused changes in the installation.
@@priotx I appreciate your prompt response. When I reached Step 5 in the instructions provided at this link, I encountered an issue where the packages numpy-1.19.4-cp36-none-manylinux2014_aarch64.whl and mediapipe-0.8.5_cuda102-cp36-none-linux_aarch64.whl are no longer supported. Thank you for your assistance. My python version is python3.8
Post inatllation, how do I confirm if the installation is successful and also how do I confirm if a program is running on the GPU or CPU on the jetson nano ?
It is running on the GPU by default, try running a basic media pipe program to check that it is working
Great tutorial! It worked!
A question: how to intercept which number (with fingers) i show with my hand?
I did not understand what you are trying to do, are you trying stop the number showing on the image
@@priotx i mean, i raise five finger and in python i want to get "5", i raise two fingers and i get "2" something like that
@@DavideOrlando1969 there are videos in RUclips about what you like to do, I will be making videos in my channel also and bring mediapipe tutorials soon.
thank you
Hi, I followed the newer steps as mentioned in your GitHub. In step 5 after I install the mediapipe-bin.zip from google drive do I need to come out of directory “cd mediapipe” and then do “cd downloads” and unzip it there? Also in “cd mediapipe-bin” has to be run inside “cd downloads”?
Do as shown in the video and install it correctly
Thanks all your videos! I love them pretty much!
Currently, I've trained a model in Google COLAB, but the trained model generated is in PYTORCH. So, I'd like to know what I have to do to run this model on the Jetson NANO. Thank you in advance.
this may help you - pytorch.org/blog/running-pytorch-models-on-jetson-nano/
I will make a video on this topic, I am busy till May 1st.
so I will the make video after May 1 st.
Is matlab can be installed in jetson nano????
🔥
I was wondering if I can do this on a virtual environmental (python 3.8) on jetson nano, because using sudo to install things means that it is not installed on my virtual environment but is actually installed on my original environment
yes you can do it on virtual environmental , but the main dependency should we installed on both root and virtual environmental , but I recommend to run it the normal way
@@priotx ok thanks !
@@TheCvin you are welcome , be connected
@@TheCvin let me check what I can do , I think we can install that as well I will try to post a new video about it
Great and easy to follow, but for some reason h5py fails to build, Ideas? I tried it on 4.6.1 and 4.6 didn't work on my Jetson B01
thank you!
Did you increase the swap before building h5py
@@priotx yes I did, I followed the instructions on GitHub
what version of ubuntu are you using 18 , 20 or smth else
@@muhammadhadi5249 18.
Does this video include building Mediapipe on the jetson nano with GPU support ? I'm trying to implement a facial detection+recognition project and I want the facial detection part of the project to run on the GPU using Mediapipe. After following the steps in the video above will I have Mediapipe built with GPU support?
Yes
Go with the github and video, as I have changed and updated on github
@@priotx Ok, I'm currently running a facail detcetion+recognition project using OpenCv and DLib, it runs at about 5fps for 2 webcams. Post Mediapipe installation, where I will be running the detection on the GPU using MediaPipe while the recognition part of the project continues to run on the CPU, Will there be a improvement in the overall frame rate?
Hello, I enjoyed the video. I'm currently using Windows, but if I dual boot Linux and run the code, will it behave the same?
This is specially for jetson nano , but if you what to try mediapipe on windows it is really easy to install and run , check other videos on youtube you will find then
@@priotx Before installing it on Jetson Nano, I wanted to test it on Linux
@@user-lo7ck7bh1i for linux or ubuntu there is a much easter method of installation on mediapipe check other videos you can find then online , after installing the same code will work on both linux and jetson nano perfectly
Melvin, Thanks for this tutorial. Unfortunately this does not allow refine_face_landmarks or complexity :( this would be great...
Glad that the video helped you
@@priotx well it didn't coz I would need distant face recognition and to do that I would need full mediapipe :( But thanks anyway ;)
awesome
How to build mediapipe with GPU support for jetson nano?
See the video and do the steps in GitHub
As the github is updated
Is it possible to compile Mediapipe on Jetson Nano?
Hi, I followed the instructıon in github link with the bin file you shared. I reached the last command in the github link but after that I don't know which commands I have to do? from which command in the video I need to continue to the end?
You may need to download the mediapipe - bin you had send to me and do old command as show in the video
@@priotx yeah, thanks a lot. It worked for me.
@@leilaghasemzade5880 ok
Hi Melvin, thank you very much for the tutorial. Can I follow the instruction for jetpack 4.5 as well? Does it work?
It was working for 4.5 and 4.4 and 4.6, but now there is a problem , SO it may not work now , I will fix the problem and make it working this week
Great Video! Does it work on Jetbot (Jetpack Version 4.4) as well?
I think it would work, I have tried it in 4.5 it works fine , so it should work in 4.4 also , try and let me know.
Are you absolutely sure that this video installs and builds Mediapipe with GPU support on the jetson nano?
Yes
Hi Melvin thanks for the explanation, I've been having trouble with the command sudo pip install opencv_contrib_python this never ends and stays in the bulding wheel for opencv-contrib-python(pyproject.toml) if you or anyone can help me, please.
Hi , I think there may be some problem in the jetpack version , which version are you using , andre on which step are you facing the problem
Hi! Can you run graphs on GPU with this installation? Thanks!
mediapipe will run on jetson graphs automatically with this install
@@priotx Will it be on CPU or GPU?
@@insoaryt gpu
Hello. I’m unable to find the exact .whl files. Do you mind posting uploading yours?
actually what problem are you facing
@@priotx I am unable to get the exact .whl files. The .whl files in the mediapipe-bin right now aren’t the same files when the video was made.
@@codaemon375 I undersand that , the github of mediapipe-bin has changed , I will update it soon in my github also , I have the solution , but need to make sure it is working fine and I will update it in my GitHub.
so please wait , I will fix it soon
@@priotx thanks!
@@codaemon375 I have upadated the github , can you try to install mediapipe now using the new way I have shown in the github and tell me it is working or not