conda create --name yolov5<\/code><\/p>\nThe following command can be used to display the various environments that have already been created.<\/p>\n
Command:<\/strong> conda info --envs<\/code><\/p>\nWhenever we will work with yolo now the environment yolov5<\/code> must be activated.<\/p>\nCommand:<\/strong> conda activate yolov5<\/code><\/p>\nNow here is the link to the GitHub project I will use.<\/p>\n
URL:<\/strong> https:\/\/github.com\/ultralytics\/yolov5\/wiki\/Train-Custom-Data<\/code><\/p>\n<\/span>Set up YoloV5 locally<\/span><\/h3>\nNow clone the yolov5 repository.<\/p>\n
Command:<\/strong> git clone https:\/\/github.com\/ultralytics\/yolov5<\/code><\/p>\nAfter cloning the repository, please change to the yolov5<\/code> folder.<\/p>\nCommand:<\/strong> cd<\/code><\/span> yolov5<\/code><\/p>\nNow install all the necessary libraries.<\/p>\n
Command:<\/strong> pip install -r requirements.txt<\/code><\/p>\nAfter everything has been installed, yolov5<\/code> is ready for use and the training data must be prepared.<\/p>\n<\/span>Create training data set of PFM-1 antipersonnel mines<\/span><\/h2>\nNow please create the training data. As described at the beginning, I used Blender to render the corresponding images of the PFM-1 mine or searched for images of this mine on the Internet. These images of the PFM-1 mines I have then automatically processed to training data and immediately create the YOLO labels for the bounding boxes. I wrote Python programs that automatically placed the transparent images as foreground images on different background images, created the labels for the bounding boxes and applied various filters to the images. So I have created well over 100,000 training datasets or generated over 100GB of data for training. How good the data really is and my approach to it remains to be seen.<\/p>\n
![\"PFM-1](\"https:\/\/ai-box.eu\/wp-content\/uploads\/2023\/05\/PFM-1-anti-personnel-landmine-training-data-overview-1024x768.png\")
<\/a>