Image augmentation techniques for Yolov5 based image segmentation

Hi Team,

I am working on an image dataset where each image contains multiple labels and frequencies of some of the labels in the dataset is very few. Thus dataset is totally imbalanced. I am planning to use Yolov5 based object detection algorithm to detect the labels in the images.
Can you please someone share some hints or guidance to solve the imbalance the dataset with respect to the labels and their annotations?
Thank you.