Hi Paul, may be upcoming lessons after completing face recognition tracking should be on detecting various types of objects and naming those objects....great going AI Guru Mr.Paul
great lesson! And, also opened up a whole new world with github! Now, I already knew about github, and use it a lot, but not to its full potential! I was not aware you could copy the files in that manner! What a huge playground you opened me up to! THank you for showing us that! Now, If I can just figure out this "GStreamer warning: Cannot query video position: Status=0, value=1, duration =-1" error so that the camera servos will track objects instead of avoid them, I'll be coming right along!
If you are grabbing and showing frames your camera is working. You can get that error even when things are all OK. The issue with the tracking must be something else in your code or connections.
@@paulmcwhorter Days of research, all way too complex for me to understand. And finally I watch jetson Xavier on Gstreamer, and finally got the answer I needed to solve my problems! You really are an amazing teacher!
Very interesting and a lot of fun lesson...stay safe in this pandemic situation and lockdown period throughout the world....Thank you for your free stuff..contribute in future if possible.
I am having a hard time copying the code from the top tech boy site. The search works. Clicking on the two page symbol causes a copy option to flash briefly rarely. Sometimes the text will be highlighted. Most times not. This is using Chromium on the Jetson Nano image. It fails on Chrome on a Windows system as well; but, less often. Thoughts?
Paul, is there any way to increase the performance of the Jetson Nano? I have installed a new up to date sd card, but there is still a lot of lag. I enjoy your videos!!!
Paul, this is the response I got back from a forum on the Jetson Nano, "Probably not. More I/O won’t help if your problem is CPU bound. I think your issues that most OpenCV code you will find out there does not use the GPU, and the CPU is not supposed to be very powerful on Tegra. You might try giving something like DeepStream a shot instead, or examining the use of TensorRT in @dusty_nv’s Jetson Inference repository". Interesting!!! Brent
Mine works but a big delay in camera. If I drop to 320x240 it is fine. All has been fine till now, not sure what is up with that. I have a hefty power supply and the fastest card in sd card tests. Fan kicks on hard too, so lots of computations going on. haha
does anyone else's camera lag when they make their window 640x480? Mine seems to only run well when i put it at 320x240. When I was debuggin it seems as if the detectMultiScale() method is the reason. Is there a way to fix this or is there just something wrong with my nano perhaps?
The lessons were for jetpack 4.3. I found poor performance on jetpack 4.4 because NVIDIA broke some of the underlying libraries. If you are on jetpack 4.4, that is probably the problem. Downgrade to jetpack 4.3
If you would like to help out, the easiest way is through Patreon, and the link is in the description of the video. I have never been to Nigeria, but have a home in Uganda, and hope to be there soon if this virus gets sorted out.
This was a very interesting video, I think that it did not detect the girl in the magazine because it was not a straight picture of her. I notice the moment I turn my head it loses me. That is just my thought I could be wrong.
I got this strange problem on the first face detection, and everything looks ok , there is no way to copy the code from toptechboy.com faces=face_cascade.detectMultiScale(gray,1.3,5) cv2.error: /home/nvidia/build_opencv/opencv/modules/objdetect/src/cascadedetect.cpp:1698: error: (-215) !empty() in function detectMultiScale I replaced the XML file, changed the path, checked the code for hours, did not help. Actually, checking the miss spelling is the thing that take most of my time here, not the learning.
Hi Paul, the jatson is expensive i understand that openCV works on raspberry pi and that the projects thst that you build can eork on the pi as well Therefore why did you choose the Nano considering you have young audience that have budget issues ?
Hi ori-ben-yehuda As you pointed out, these lessons, with slight modifications, do work on the raspberry pi. I've been following along on a raspberry pi 3 and up until now, including this lesson, everything works fine. However, it is a little slower, and as we progress into more gpu intensive lessons, I think the raspberry pi will be really slow. The Jetson Nano is just a lot faster because of the 128 cores. Also, I am guessing that future lessons will require downloads that only work on the Jetson Nano and not on the raspberrypi since the Nano is geared toward Artificial Intelligence and the raspberry pi is less so. But the young audience you mention can and should use the raspberry pi as far as it will take them.
Love it, on to today's homework. Such dedication by the way, recording a video in a bleach cleaned room while being allergic, thanks for this.
Paul you literally saved my life with your videos, you're a beautiful mind.
Wow, Now I can detect my face using netson nano. Very fascinated to do facial recognition and Thanks a lot for these videos sir.
Hi Paul, may be upcoming lessons after completing face recognition tracking should be on detecting various types of objects and naming those objects....great going AI Guru Mr.Paul
Hi Paul, Thanks, great lesson. The little 2g is still in there! Used sg90 servos on tracking too.
great lesson! And, also opened up a whole new world with github! Now, I already knew about github, and use it a lot, but not to its full potential! I was not aware you could copy the files in that manner! What a huge playground you opened me up to! THank you for showing us that!
Now, If I can just figure out this "GStreamer warning: Cannot query video position: Status=0, value=1, duration =-1" error so that the camera servos will track objects instead of avoid them, I'll be coming right along!
If you are grabbing and showing frames your camera is working. You can get that error even when things are all OK. The issue with the tracking must be something else in your code or connections.
@@paulmcwhorter Days of research, all way too complex for me to understand. And finally I watch jetson Xavier on Gstreamer, and finally got the answer I needed to solve my problems! You really are an amazing teacher!
This was a FANTASTIC LESSON Paul. Bravo, keep going !!
Super that one of your students was on the cover of QST. Going right now to dig my copy out of my library!
Watched it 'till the end! I loved it and I am really excited about what am I going to learn in future lessons.
Great lesson Paul ! You are making this stuff easy!
This was great...I'm excited about where this leads.
Count down starts !!!
i can hardly wait
Thanks! Great lesson! As allways... now I’m going to do the homework!
Fantastic lesson as usual. keep going ! Stay safe !
Brilliant, awaiting for such lessons 👍
Paul it will be great if you teach us how to track a face after the camera recognises a face and move the servo motors
Really enjoyed this lesson Paul - Thanks
Hi Paul, this was a fun lesson! Still waiting for my servos...
Another great lesson. I made it.
Another interesting video Paul! but what could be the Coronavirus Secret Cure? 😂
Very interesting and a lot of fun lesson...stay safe in this pandemic situation and lockdown period throughout the world....Thank you for your free stuff..contribute in future if possible.
sorry the path was the problem, works now :)
Awesome. Amazing. Thanks for lesson ;)
Paul, please remember to toggle off the file list in order to show more of the code when the list is not longer require. Thanks.
Hi Paul, i wanted to know if you d consider making C++ lessons eventually on the raspberry pi?
Hi, Nice lesson. How ever I think you elaborate more of what is a cascade. What is this xml . what its content ? Thank you for your wonderful lessons
This lesson beat my Pi4 (4gb) into submission! Everything is working perfectly though on my Nano including 2 eyes.
Great video, this was a breeze. I am concerned about the lagginess of my setup. Any pointers? I am using an Arducam 477 setup.
I am having a hard time copying the code from the top tech boy site. The search works. Clicking on the two page symbol causes a copy option to flash briefly rarely. Sometimes the text will be highlighted. Most times not. This is using Chromium on the Jetson Nano image. It fails on Chrome on a Windows system as well; but, less often.
Thoughts?
Hi Paul, the Jetson Nano seems to run very slow when running this program, even though my code is identical. Any ideas as to why this might be?
Same issue with me, did u solve this problem?
Hi paul, i need to detect one object that i decide. how can i train my model? from where can i learn it?
Paul, is there any way to increase the performance of the Jetson Nano? I have installed a new up to date sd card, but there is still a lot of lag.
I enjoy your videos!!!
Paul, this is the response I got back from a forum on the Jetson Nano,
"Probably not. More I/O won’t help if your problem is CPU bound. I think your issues that most OpenCV code you will find out there does not use the GPU, and the CPU is not supposed to be very powerful on Tegra.
You might try giving something like DeepStream a shot instead, or examining the use of TensorRT in @dusty_nv’s Jetson Inference repository".
Interesting!!!
Brent
Mine works but a big delay in camera. If I drop to 320x240 it is fine. All has been fine till now, not sure what is up with that. I have a hefty power supply and the fastest card in sd card tests. Fan kicks on hard too, so lots of computations going on. haha
did you ever find out how to fix that problem? I am having the same issue
Is it possible improve the program? Haarcascades ins't accurate, depending on the possition it doesn't detect.
does anyone else's camera lag when they make their window 640x480? Mine seems to only run well when i put it at 320x240. When I was debuggin it seems as if the detectMultiScale() method is the reason. Is there a way to fix this or is there just something wrong with my nano perhaps?
What camera are you using. If you are using a web cam make sure you start it with '/dev/video1' not just a '1' if you are on jetpack 4.3 or 4.4
Paul McWhorter im just using the picam can I still do this?
The lessons were for jetpack 4.3. I found poor performance on jetpack 4.4 because NVIDIA broke some of the underlying libraries. If you are on jetpack 4.4, that is probably the problem. Downgrade to jetpack 4.3
Paul McWhorter ok thank you
there is actually still delay perhaps theres something wrong internally
If there were only three people like Paul McWhorter in Nigeria, our country would have been very different. Could you guide me on how to donate sir
If you would like to help out, the easiest way is through Patreon, and the link is in the description of the video. I have never been to Nigeria, but have a home in Uganda, and hope to be there soon if this virus gets sorted out.
Hi sir can you teach CV with medical images
mission done, eyes are easier to being detected, maybe need to fine tune the parameters
This was a very interesting video, I think that it did not detect the girl in the magazine because it was not a straight picture of her. I notice the moment I turn my head it loses me. That is just my thought I could be wrong.
I got this strange problem on the first face detection, and everything looks ok , there is no way to copy the code from toptechboy.com
faces=face_cascade.detectMultiScale(gray,1.3,5)
cv2.error: /home/nvidia/build_opencv/opencv/modules/objdetect/src/cascadedetect.cpp:1698: error: (-215) !empty() in function detectMultiScale
I replaced the XML file, changed the path, checked the code for hours, did not help.
Actually, checking the miss spelling is the thing that take most of my time here, not the learning.
Hi Paul, the jatson is expensive i understand that openCV works on raspberry pi and that the projects thst that you build can eork on the pi as well
Therefore why did you choose the Nano considering you have young audience that have budget issues ?
Hi ori-ben-yehuda As you pointed out, these lessons, with slight modifications, do work on the raspberry pi. I've been following along on a raspberry pi 3 and up until now, including this lesson, everything works fine. However, it is a little slower, and as we progress into more gpu intensive lessons, I think the raspberry pi will be really slow. The Jetson Nano is just a lot faster because of the 128 cores. Also, I am guessing that future lessons will require downloads that only work on the Jetson Nano and not on the raspberrypi since the Nano is geared toward Artificial Intelligence and the raspberry pi is less so. But the young audience you mention can and should use the raspberry pi as far as it will take them.