Crop and Save YOLOv4 Object Detections | Custom YOLOv4 Functions with TensorFlow

Learn how to code your very own Custom Functions to work with YOLOv4 Object Detections! In this video I will walk-through a custom function I have created to crop and save YOLOv4 object detections as new images. This tutorial can be implemented using YOLOv4, YOLOv4-tiny, YOLOv3, or YOLOv3-tiny. The commands can be run using TensorFlow 2.0, TensorFlow Lite or TensorRT models on images, video and webcam!

#yolov4 #objectdetection #tensorflow

This video covers the implementation of YOLOv4 using TensorFlow and how to add a command line argument in order to activate the custom function for cropping of detections.

Video Breakdown:
1. Cloning or Downloading the Code for the Tutorial
2. Running YOLOv4 Object Detector on Images using the custom crop flag
3. Show where the cropped new images are saved within the repository.
4. Running YOLOv4 Object Detector on Video using the custom crop flag
5. Explaining the code for the custom cropping function and how to edit it

--------------------------Resources---------------------------

If you enjoyed the video, toss it a like! 👍

Thanks so much for watching!
- The AI Guy

[TheAIGuy]

As always I hope you enjoy! Let me know if you run into any issues and let me know which videos you want to see next. :)

[joooooojo]

thank you so much! your video is very clear to follow!
I am wondering if you would make another video for checking the accuracy of the detection(detecting another labelled dataset that is not used in the training)
thanks again!

[victoriapark]

Thank you so much and please make a tutorial about pose estimation. There are some in youtube but they don't work. Yours are always working perfectly.

[rashikrahman]

I faced an issue where there was no crop image to save when I tried detect_video.py so it raised an error. Thus in functions.py line 57, I modified the code a little bit

try:
cv2.imwrite(img_path, cropped_img)
except:

print('No Image')
continue

Now it doesn't throw errors and completes the whole video. Thought it may be helpful for someone having the same issues.

Love your work. Keep on making tutorials like this. You don't know how much this means to us. Learned a lot, thanks...

[colinfredynand]

I’m getting an error and shows
ValueError: cannot reshape array of size 2126904 into shape (1024, 512, 3, 3)
I’m using a custom yolov4 trained weight with obj.names containing only a single class
Correction made in config.py
Still showing this error, tried with a yolov3 custom trained weight and also same situation
All done in colab

[dipteshchakraborty]

Thanks AI guy, for all the videos. I have done the same using cv2. Thanks for your inspiration

[isabellafukushima]

I have learned so much from you these past weeks, thank you so much for your hard work and the great content!

[kvotheosem-sangue]

Wow thanks for an official solution, was doing it by changing those .c files :P

[jokosusilo]

Thanksss, I learn from completely zero from you. Now, I modify your code and make exactly the same but I combine it with DeepSORT and only crop the object at its highest confidence score. Next, for your tutorial, would you please create about photogrametry?

[vigneshmahalingam]

Hey @AIGuy,
Your work is Great!! Your videos are really clear, very neatly sequenced and well explained.

I would like to place a question if you will be able to get into the evaluation part and make a few videos on them.

1. How is mAP calculated?
2. How is FPS calculated?
3. How to arrive at the similar yolo paper results documentation interms of mAP, FPS, etc.

Thank you.
Hoping to hear from you on this query.

[michaelmutekeri]

I am a fan, waiting for the next video. Thanx mate

[kshitizsharma]

Hello sir i just want to say a huge thank you to you for the content you are providing i really wish you achieve great heights in your life just amazing content sir thank you so much

[joshpauline]

Mate you're a king, these videos have helped me heaps

[Sylvanas]

Thank you so much for this video - this is incredibly helpful.

[necbranduc]

Nice video, I had to do something similar at work! If you only posted this video 2 weeks ago, I would have saved about 4 hours of work. Btw, update your Chrome browser😅

[Sagar_Tachtode_]

was searching for this for a long time. thanks, Man!!!

[drhope]

Hi thank you so much for your vedii pleas how can i make classification with thos crop images

[ludensfield]

Great work! This is very useful.
Following your previous video on customizing YOLOv3 using TensorFlow, how can I achieve something like this as well?

[kevinmakumbe]

Awesome video, how can i crop all the text in an image document

[alankarshukla]

This is good one. Is there any video based on SSD you made.