Model recognizing numbers in the hand sign with YOLOv4

YOLOv4 is a powerful and efficient object detection model that strikes a balance between speed and accuracy. Its use of unique features and bag of freebies techniques during training allows it to perform excellently in real-time object detection tasks. YOLOv4 can be trained and used by anyone with a conventional GPU, making it accessible and practical for a wide range of applications.

Training with Google Colab

This model training was done through Google colab. So what is Google Colab? Google Colab provides a free usable training environment. By providing GPU and TPU available in this training environment, it provides a faster model training. The model training took about 4 hours with 950 photos. This is a very small number of photos for a normal model training. The model's dataset was taken from "https://www.kaggle.com/"

Labeling of the dataset

The labeling of the dataset was done with the "https://roboflow.com/" site. First of all, the received dataset was divided into folders by classifying them according to numbers. Then the folders were labeled one by one as 6 classes. The cfg file was also edited accordingly.The model was successfully trained after these procedures.

Here is the training process in Google Colab

You can access the Google Colab code of the Model Training here: "https://colab.research.google.com/drive/1WZUsela94U6TCqd-e2XcUdltcIp5Err5?ouid=100280463691588925641"

Start phase

After the labeling process, I trained the model on Google Colab and then I wrote a Python code to detect and write down the significant numbers from the hand gestures made to the camera. Python code coded on Jupyter Notebook.

Requirements:

Python (3.x)

OpenCV (cv2)

Google Colab Environment

Data set

Contributions:

I am waiting for the feedback of my project, I hope it worked for you. I wish you many YOLO days :D

Author:

Buse Berren ÜNAL