Counting Objects Using YOLOv4 Object Detection | 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 how to run an object counting app 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 will cover two custom functions that have been created with YOLOv4. The first being for counting objects within images and video, and the second being a custom flag to show detailed information on YOLOv4 detections (class, confidence score, bounding box coordinates).

Follow the tutorial to be able to convert YOLO style Darknet weights into saved TensorFlow models, and run these models. Using easy command line flags to enable custom functions like object counting. Take advantage of YOLOv4 as a TensorFlow Lite model, it's small lightweight size makes it perfect for mobile and edge devices such as a raspberry pi. Looking to harness the full powers of a GPU? Then run YOLOv4 with TensorFlow TensorRT to increase performance by up to 8x times.

In this video I cover:
1. Show Custom Functions Repository and Explain Object Counting
2. Cloning or Downloading the Code
3. Downloading and Converting YOLOv4 Weights into a saved TensorFlow Model
4. Enabling Object Counting while performing YOLOv4 Object Detections with TensorFlow on images, video and webcam
8. Printing Detailed Information About Object Detections to command prompt
9. Stepping through the code to create custom functions

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

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

Thanks so much for watching!
- The AI Guy

[TheAIGuy]

Let me know which Custom Functions you want me to build next here in the comments! Cheers :)

[akshayjaryal]

Thanks for the informative learning & Coding !!

[ahmadyoussef]

Dude you are amazing, and i am pissed cause another channel took your stuff and used it as the easiest way to install yoloV4 and they have more subscribers but u got the videos and the repo and everything, you are amazing keep it up and thank you so much for the help here

[SivaShankarsss]

Sir, you are great...Really helpful to learn yolov4

[zeljko]

Amazing video, code is clear and easy to understand. Idea, maybe make a passing lines and count objects that passes that line from same direction? i.e. in and out

[himesh_]

Great add on would be, draw line and count number of people crossing the line.

[zhihaozhou]

amazing as always!!! would you plz upload a Colab tutorial of this? and use pre-trained weights to detect one class? benefit a lot from your channel !!

[vijayanand]

Amazing Tutorial🤩. Looking for more amazing videos!

[annielammy]

Thank you for the tutorials. Can you apply this process to the tensorflow colab as well?

[chhandakpatra]

Please make a speed detection of objects(for cars trucks bikes etc) and line crossing counter on road that could be helpful to do tracking on vehicles which breaks traffic rules

[akshay]

Hi, I want to count th enumber of objects in a continuous video feed of camera and store it in real time. How do i do that? for example, if theres a rowof straawberries, as i move the camera, i want to count the number of strawberries it has detected without counting the same strawberry twice. Thanks in advance.

[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

[ahmadadib]

Can we detect multiple images in 1 folder, not doing it with one by one images?

[barnowldroneservices]

Great stuff here! Highly recommended. Thank you

[ZkyMaster]

Any ideas about why I got the warning during darknet to TF conversion when I use Tensorflow GPU 2.6.0? The warning is like this: WARNING:TensorFlow:Compiled the loaded model, but the compiled metrics have yet to be be empty until you train or evaluate the model. After that when I try to do inference multiple images, the bbox result only appears only on the first image, and not the rest of them. Same with the video, only inference the first frame. Any help would be appreciated 🙏.

[TheNydeon]

Great video once again and very informative ! The code walkthrough was very nice.
About the video detection console output, would it be easy to add when an object was found in the video? For example the output would be : Object found at 01:23:433 : dog ...

Thanks a lot :)

[luhaoming]

Hi there,
First of all, thx for your work. It is a nice program.
I have a question about the counter in the project.
If I want to display the total number of detections and the number of classification detections, instead of showing the classification number and the total number of detections in the current frame. How do I change the function.py?
The first time, I dele the 'else' in the count. But only the items detected in the current frame can be displayed, and the sum function cannot be realized.
Can u give me any advice?

[drewgibson]

Love it!! and I love the issues system!!

[heinoplaatjies]

great video, have you done something similar in Yolov5

[willmcgowan]

Great video. Has anyone else having problems where the detection's only work for the first frame?