YOLOv4 is “technically” the latest state of the art object detector. And that is because some might argue the same in favour of YOLOv5, but since that is highly contested at the moment, let’s just go with YOLOv4 for now. I’m not going to go into what this technology is, but I’m going to dive straight into the method to create your own customised dataset that can be used for object detection with YOLOv4 and Darknet using Google’s Open Images Dataset which has 600 classes of image data that you can choose from.
A big shout-out to Vittorio Mazzi for building an amazing toolkit for downloading and labelling images from the Open Images Dataset. And special thanks to The AI Guy for creating a great way to label those images using annotations in YOLOv4 format. This article is inspired from their creation.
Check out their github repos for detailed explanation on the same:
EscVM/OIDv4_ToolKit
theAIGuysCode/OIDv4_ToolKit
Note : You need to have Python downloaded in your local machine for the steps mentioned below to work. So if you don’t already have it, download it from here.
Steps to create your own dataset :
- Step 1 : To start off, go to your command prompt or terminal and type the following command to clone the github repo of theAIGuysCode/OIDv4_ToolKit into your local machine in your preferred folder.
git clone https://github.com/theAIGuysCode/OIDv4_ToolKit.git
You could also just download the zip file from the repo directly.
- Step 2: From your terminal, go inside the folder ‘OIDv4_ToolKit’ and type in:
pip install -r requirements.txt
This will get you all the libraries that would be necessary for implementing this project.
#yolov4 #deep-learning #darknet #deep learning
