Running a little behind in trying to keep up with the weekly lessons, but I think this may be my favorite so far! But I might have to eat my words on the next lesson. Time will tell! Wonderful stuff Paul!! And I know that now that school is back in session it has thrown a kink into your video production schedule. Along with the sale of your place in TX.
Really getting interested I was following dusty nv pages but as I realised nothing I can to with it but you absolutely made it interesting and useful...❤
This a crucial lesson . I somehow skipped it and tried to mix OpenCV commands with Jetson detection without the format conversions taught here. The result, lots of obscure errors that even Google could not help me with. Lesson 53 should be in print in bold and red.
Hi Paul, The font size is ok for reading. I've used the PiCam and I did not forgot about the appsink drop=true setting in the camSet to get rid of that delay. I'm testing both jetson.utils and openCV environments, the fps difference is 18fps vs 14.5 fps (with the rectangles and labels shown). Also the issue with jetson.utils window resize is now fixed, does not crop the picture anymore.
Hi Paul, thanks for the amazins series. I have a problem with todays lesson : I get an error with the X11 system, the programm does not allow the creation of the display : Do you have any ideas ? [OpenGL] glDisplay -- X screen 0 resolution: 1920x1080 [OpenGL] failed to create X11 Window. jetson.utils -- PyDisplay_Dealloc()
I also got the failed to create glDisplay device. I moved the import cv2 line before import jetson.inference and it now works fine. Go figure. Importing in this order works: import cv2 import jetson.inference import jetson.utils import time Importing in the order shown in the video fails with the slightly more recent libraries I'm running. This import order fails: import jetson.inference import jetson.utils import time import cv2 With this error message [OpenGL] failed to create X11 Window. [OpenGL] failed to create OpenGL window Traceback (most recent call last): File "/home/cat/Desktop/pyPro/faceRecognizer/Nvidia/deepLearning5ObjectDetection.py", line 13, in display=jetson.utils.glDisplay() Exception: jetson.utils -- failed to create glDisplay device
Great lesson Paul, Really like how python handles data structures.... no primitive building!!!! Had to approach this a bit different on the 2g do not go to cv2 until after detections and for loop under that. Just not enough memory for doing it all the way also went with 'ssd-mobilenet-v1' to save space. Not that it uses it often but put a fan on it also. With web cam or pi cam getting 20fps. Running JP 4.5. Thanks!!!!
Makes me want to try jetpack 4.5. I really need to go back and revisit these videos. Lots has changed. Putting together a series like this becomes a challenge as so much changes so quickly. It is almost like I need to start the series by saying which jetpack i am on and telling people to use that one, even if later and greater ones are out.
Hi Paul, how could you add your own custom images (not faces or people, but objects) to improve the, let’s say ssd mobilenet dataset, and still move a motor with it like your previous lesson. I followed Dusty’s tutorial on training your own object detector but like you said, no freedom whatsoever on integrating to a robot.
Hi Paul, Are you going to get into more than just face recognition with the Jetson Nano? It would be nice to be able to detect/classify objects like planes and cars too!
Hi Thomas, if you go to Dusty's github page that contains the tutorials the Paul is working with (github.com/dusty-nv/jetson-inference) there are pre-trained object detection models available for planes, cars and more! Look at the models with the COCO classes (includes airplanes and cars) and the other ones available in the model library.
@@GOBish23 Hi Gabi, Thanks for the suggesstion, much appreciated. I already have a Jetson nano set up with Dusty's version of the jetson-inference software. As you must be aware the jetson-inference does not make use of openCV where as Paul's does. The Tensorflow/Keras and PyTorch used in the jetson-inference does not have an object detection box capability and uses semantic segmentation (github.com/dusty-nv/jetson-inference/blob/master/docs/segnet-console-2.md) to identify objects. Dusty has stated that he has tried to create object boxes but with no success.
Great Video, i've learn from you more than any of my teacher at school. But i got a question: Normally on my PC, once i installed everything correctly, when i type "cv2. " there will be the list of suggestions appear. However as i can see from your videos and trying on my Jetson Nano, there are no any suggestions for the cv2 library. This auto suggest feature of Vscode worked perfectly if i import and call out a function from other python files. I hope you have a answer for this.
If i were to use a video file input instead of a camera stream, how would i change the code? I tried using videosource from jetson utils but the program does not run properly with the code you wrote for a live camera.
yes we do, I'm not sure why you haven't looked into what you can pass glDisplay. I love your videos, but I can not see going back to OpenCV which will take the load off of the GPU and put it on the CPU. Man I sure hope you know this by now. As well by going OpenCV you are taking a big FPS hit.
Quite exciting. Breaking the data down and finding what item ID# goes to what item, cool stuff. Thanks, Paul.
thank you very much Paul; I'm part of the young generation and I am infinitely grateful to you for the content you share
Running a little behind in trying to keep up with the weekly lessons, but I think this may be my favorite so far! But I might have to eat my words on the next lesson. Time will tell! Wonderful stuff Paul!! And I know that now that school is back in session it has thrown a kink into your video production schedule. Along with the sale of your place in TX.
Hi Paul, this was a really cool lesson. Thank you very much. Laurent(FR)
Really getting interested I was following dusty nv pages but as I realised nothing I can to with it but you absolutely made it interesting and useful...❤
All respect for detailed and prof. content you are providing
god bless. So far i ve watched all the videos on the playlist
I'm still here and loving your videos. :)
This a crucial lesson . I somehow skipped it and tried to mix OpenCV commands with Jetson detection without the format conversions taught here. The result, lots of obscure errors that even Google could not help me with. Lesson 53 should be in print in bold and red.
Hi Paul,
The font size is ok for reading.
I've used the PiCam and I did not forgot about the appsink drop=true setting in the camSet to get rid of that delay.
I'm testing both jetson.utils and openCV environments, the fps difference is 18fps vs 14.5 fps (with the rectangles and labels shown). Also the issue with jetson.utils window resize is now fixed, does not crop the picture anymore.
Hi Paul,
thanks for the amazins series. I have a problem with todays lesson : I get an error with the X11 system, the programm does not allow the creation of the display : Do you have any ideas ?
[OpenGL] glDisplay -- X screen 0 resolution: 1920x1080
[OpenGL] failed to create X11 Window.
jetson.utils -- PyDisplay_Dealloc()
Hello, just import cv2 first and then import jetson.inference, jetson.utilis. It worked for me. I also had the same error.
I had the same issue and solved
Are you using raspberry pi camera?
Then try commenting out
# display = jetson.utils.glDisplay()
I also got the failed to create glDisplay device. I moved the import cv2 line before import jetson.inference and it now works fine. Go figure.
Importing in this order works:
import cv2
import jetson.inference
import jetson.utils
import time
Importing in the order shown in the video fails with the slightly more recent libraries I'm running. This import order fails:
import jetson.inference
import jetson.utils
import time
import cv2
With this error message
[OpenGL] failed to create X11 Window.
[OpenGL] failed to create OpenGL window
Traceback (most recent call last):
File "/home/cat/Desktop/pyPro/faceRecognizer/Nvidia/deepLearning5ObjectDetection.py", line 13, in
display=jetson.utils.glDisplay()
Exception: jetson.utils -- failed to create glDisplay device
Yeah, I am using Jetpack 4.5 and also have this issue. This solve the problem, thank you guys!
much awaited lesson
Great lesson Paul
Great lesson Paul, Really like how python handles data structures.... no primitive building!!!! Had to approach this a bit different on the 2g do not go to cv2 until after detections and for loop under that. Just not enough memory for doing it all the way also went with 'ssd-mobilenet-v1' to save space. Not that it uses it often but put a fan on it also. With web cam or pi cam getting 20fps. Running JP 4.5. Thanks!!!!
Makes me want to try jetpack 4.5. I really need to go back and revisit these videos. Lots has changed. Putting together a series like this becomes a challenge as so much changes so quickly. It is almost like I need to start the series by saying which jetpack i am on and telling people to use that one, even if later and greater ones are out.
Hi Paul, how could you add your own custom images (not faces or people, but objects) to improve the, let’s say ssd mobilenet dataset, and still move a motor with it like your previous lesson. I followed Dusty’s tutorial on training your own object detector but like you said, no freedom whatsoever on integrating to a robot.
Hi Paul, Are you going to get into more than just face recognition with the Jetson Nano? It would be nice to be able to detect/classify objects like planes and cars too!
Hi Thomas, if you go to Dusty's github page that contains the tutorials the Paul is working with (github.com/dusty-nv/jetson-inference) there are pre-trained object detection models available for planes, cars and more! Look at the models with the COCO classes (includes airplanes and cars) and the other ones available in the model library.
@@GOBish23 Hi Gabi, Thanks for the suggesstion, much appreciated. I already have a Jetson nano set up with Dusty's version of the jetson-inference software. As you must be aware the jetson-inference does not make use of openCV where as Paul's does. The Tensorflow/Keras and PyTorch used in the jetson-inference does not have an object detection box capability and uses semantic segmentation (github.com/dusty-nv/jetson-inference/blob/master/docs/segnet-console-2.md) to identify objects. Dusty has stated that he has tried to create object boxes but with no success.
Great Video, i've learn from you more than any of my teacher at school.
But i got a question:
Normally on my PC, once i installed everything correctly, when i type "cv2. " there will be the list of suggestions appear. However as i can see from your videos and trying on my Jetson Nano, there are no any suggestions for the cv2 library. This auto suggest feature of Vscode worked perfectly if i import and call out a function from other python files.
I hope you have a answer for this.
Sir .... Thank you so much... Benifited a lot by your videos sir.... Please do more Raspberry pi projects sir.....
This is amazing .. thanks for your efforts and time!
Hi Paul, great videos! If I wanted to use an IP camera (RTSP) stream, what would the Cam= variable be?
Great lesson. Great channel
Thanks for the lesson. Size is good enough i think.
I have a question about how to connect the audio out input in jetson nano, for zoom.
Yikes, I just went thorough your arduino series, and I forgot how to python. Great Series BTW.
Just errors out for me.
gst_is_element' failed
line 30
height=img.shape[0]
attributeerror: 'nonetype' object has no attribute 'shape'
Me too. I will let you know if I find a solution.
Mee tooo
is there a way to use servo control inside the docker container for object tracking
If i were to use a video file input instead of a camera stream, how would i change the code? I tried using videosource from jetson utils but the program does not run properly with the code you wrote for a live camera.
How do you display confidence values as well?
Shazam! Great lesson sir
This is working great on my webcam, on my pi cam it crashes. When removing cam.set(cv2.CAP_PROP_FRAMEWIDTH, dispW). and Height. it works.
Great lesson. Can't observe the result of Visual studio code sir..otherwise its OK. Thank you.
yes we do, I'm not sure why you haven't looked into what you can pass glDisplay. I love your videos, but I can not see going back to OpenCV which will take the load off of the GPU and put it on the CPU. Man I sure hope you know this by now. As well by going OpenCV you are taking a big FPS hit.
Can you do a video for vnc for the jetson nano?
How to connect a mic in jetson nano
Font size is optimal
34:00