Note: I've updated the code for this video to work with TensorFlow 2.0. We've changed to yolov3-tf2 for the YOLO library. Changes to the video as a result of this change are very minor, so I am not yet rerecording. As the description says, new code is here: github.com/jeffheaton/t81_558_deep_learning/blob/master/t81_558_class_06_5_yolo.ipynb
Thanks Jeff for an excellent playlist. Does the code in the above link need any modification, I tried to create directories in Colab to download the yolo files but it is not working. Thank you
Thank you Jeff Heaton for all your excellent lectures about python and deep learning, I would like to do class 06_05_yolo, but this given github link does not have the code which is shown in the video, the code is the older version of this video, can you please update it ?
Thank you sir, for sharing your expertise. I'm going to train a model to recognize cigarette magots as a simple project. I want to use it in a simple autonomous magots collecting robot as a useful staring point, before adding more sorts of littering. Students, feel free to pick up this idea.
Hello Mr. Jeff Heaton I get this error in the before last code ==> module 'tensorflow.python.distribute.distribute_lib' has no attribute 'StrategyV1' Do have any idea ?
Is the training input to YOLO images with or without bounding boxes or labels? If without, is YOLO predicting bounding box location in an unsupervised fashion? For transfer learning for more specific detection tasks would your new data need bounding boxes or labels of any kind?
Hi, our brilliant teacher. I have run the code provided by you locally on my own machine, which makes me very happy. I am curious about how can I draw a rectangle around the detected object just like displayed at the beginning of this tutorial?
TH-camr Mark Jay has a great tutorial about this called "Image Detection with YOLO-v2 (pt.2) Process Image in Python + openCV". The video links to his Github repo and you'll easily find the code there (inputs 20 and 22). 😊 The idea is the tfnet.return_predict(img) returns a list of dictionaries (one dictionary per bounding box.) Each dictionary has the coordinates and label for the bounding box. You can extract these values them plot them on your image.
I've just realized that Jeff also has the code for drawing the bounding boxes on the image. It's in the github page listed in the description for this video. 😊
Is there a tutorial about retraining YOLOv3 or v4 with custom dataset via transfer learning? In other words, I'd like to transfer the weights learned and not training the model from scratch.
oh no! how to generate bounded box images as output? Is it possible to to generate video with bounding boxes just like you have shown ..? how to do that?
I managed to get the cell “Installing YoloV3-TF2" to run by installing GIT on my PC, however now the cell on" Transferring Weights" fails saying "No module named 'cv2'". So, I tried to install OpenCV but that seems to be more difficult than it first appeared (ie pip install opencv-python did not work). I was able to run the code in CoLab by creating the necessary folders in my Google drive. It's all a bit of a shame I can't get OpenCV to work locally as everything has been fine up to this point and this is the most interesting part! In case anyone else has this problem, with help from a friend, my program now runs. The issue is to correctly install OpenCV in the right base / virtual environment so that import cv2 gives no error. To do this, after uninstalling OpenCV, I did the following: • conda activate tensorflow • Then switched to the folder I am working in (t81_588_deep_learning-master) • Installed OpenCV (pip install opencv-python) • Started Python and did the test - ie import CV2 and printing version • Then started jupyter notebook and tested if import CV2 gave no error
15:00 please tell me how can I upload my own image and test this I tried uploading image to Google drive and created link and then I copy pasted that link in url variable but it's not working please tell how to do!!
It can be tricky, you have to have exactly the right weights aligned to the right config file. Its not as clear as it could be on the YOLO site, it took me a few ties/combinations to make it work.
@@HeatonResearch Thanks for your reply. I will try various combinations by myself. I appreciate your tutorial, which let me learn a lot of new knowledge. Have a nice day.
@@beizhou2488 If it does not work let me know and I can probably add exact links to a combination that worked for me. I don't think I am allowed to post the actual files.
@@HeatonResearch Yeah. You may not be allowed to put the actual files. You can put the link at here or upload the file to Google Drive and then put the sharing link here. Thank you very much.
Note: I've updated the code for this video to work with TensorFlow 2.0. We've changed to yolov3-tf2 for the YOLO library. Changes to the video as a result of this change are very minor, so I am not yet rerecording. As the description says, new code is here: github.com/jeffheaton/t81_558_deep_learning/blob/master/t81_558_class_06_5_yolo.ipynb
this given github link does not have the code which is shown in the video, please provide that one for running it in colab
Thanks Jeff for an excellent playlist. Does the code in the above link need any modification, I tried to create directories in Colab to download the yolo files but it is not working. Thank you
Plz sir a need some helps if can a contact you on WhatsApp or email plz
Thank you Jeff Heaton for all your excellent lectures about python and deep learning, I would like to do class 06_05_yolo, but this given github link does not have the code which is shown in the video, the code is the older version of this video, can you please update it ?
Could you please explain temporal CNN, the architecture and stuff?
Great Video Mr. Jeff Heaton Thank you
Thank you sir, for sharing your expertise.
I'm going to train a model to recognize cigarette magots as a simple project.
I want to use it in a simple autonomous magots collecting robot as a useful staring point, before adding more sorts of littering.
Students, feel free to pick up this idea.
Best of luck! Sounds like an interesting project.
Hello Mr. Jeff Heaton
I get this error in the before last code ==> module 'tensorflow.python.distribute.distribute_lib' has no attribute 'StrategyV1'
Do have any idea ?
"House"? 8:16 Every time I see one of those 4-legged beasts I'm now going to call it a "House"... I've been properly trained! ;-)
Nice tutorial, How do I use my own trained classifier with yolo?
Is the training input to YOLO images with or without bounding boxes or labels? If without, is YOLO predicting bounding box location in an unsupervised fashion? For transfer learning for more specific detection tasks would your new data need bounding boxes or labels of any kind?
Hi, our brilliant teacher. I have run the code provided by you locally on my own machine, which makes me very happy. I am curious about how can I draw a rectangle around the detected object just like displayed at the beginning of this tutorial?
TH-camr Mark Jay has a great tutorial about this called "Image Detection with YOLO-v2 (pt.2) Process Image in Python + openCV". The video links to his Github repo and you'll easily find the code there (inputs 20 and 22). 😊
The idea is the tfnet.return_predict(img) returns a list of dictionaries (one dictionary per bounding box.) Each dictionary has the coordinates and label for the bounding box. You can extract these values them plot them on your image.
I've just realized that Jeff also has the code for drawing the bounding boxes on the image. It's in the github page listed in the description for this video. 😊
Is there a tutorial about retraining YOLOv3 or v4 with custom dataset via transfer learning? In other words, I'd like to transfer the weights learned and not training the model from scratch.
Jeff , do you do privet classes?
Not at this point.
Great Tutorial, have you ever done a tutorial on training tiny Yolo?
oh no! how to generate bounded box images as output? Is it possible to to generate video with bounding boxes just like you have shown ..? how to do that?
Jeff, any live class I could attend
It is all online for Fall 2020.
So good 🙈😍😍😍
I managed to get the cell “Installing YoloV3-TF2" to run by installing GIT on my PC, however now the cell on" Transferring Weights" fails saying "No module named 'cv2'". So, I tried to install OpenCV but that seems to be more difficult than it first appeared (ie pip install opencv-python did not work). I was able to run the code in CoLab by creating the necessary folders in my Google drive. It's all a bit of a shame I can't get OpenCV to work locally as everything has been fine up to this point and this is the most interesting part!
In case anyone else has this problem, with help from a friend, my program now runs. The issue is to correctly install OpenCV in the right base / virtual environment so that import cv2 gives no error. To do this, after uninstalling OpenCV, I did the following:
• conda activate tensorflow
• Then switched to the folder I am working in (t81_588_deep_learning-master)
• Installed OpenCV (pip install opencv-python)
• Started Python and did the test - ie import CV2 and printing version
• Then started jupyter notebook and tested if import CV2 gave no error
15:00 please tell me how can I upload my own image and test this
I tried uploading image to Google drive and created link and then I copy pasted that link in url variable but it's not working please tell how to do!!
U should try by mounting your google drive with colab file
I tried to use yolov3 and downloaded the yolov3.cfg, yolov3.weights accordingly. But it does not work for me.
It can be tricky, you have to have exactly the right weights aligned to the right config file. Its not as clear as it could be on the YOLO site, it took me a few ties/combinations to make it work.
@@HeatonResearch Thanks for your reply. I will try various combinations by myself. I appreciate your tutorial, which let me learn a lot of new knowledge. Have a nice day.
@@beizhou2488 If it does not work let me know and I can probably add exact links to a combination that worked for me. I don't think I am allowed to post the actual files.
@@HeatonResearch Yeah. You may not be allowed to put the actual files. You can put the link at here or upload the file to Google Drive and then put the sharing link here. Thank you very much.
another video please using opencv for yolov3.
It would be cool if it didnt finish so quickly but ok.