An inclusive study of LISC and Kaggle White blood cell Classification using YoloV5 Variants.

White blood cell (WBC) is crucial in human being for their immunity, genetic association, clinical significance , but it plays a vital role in medical diagnosis, particularly in the cure and real time monitoring of a variety of blood diseases. Any Infection in the body can be easily detected by doing count of WBC sub types. Traditional WBC cataloguing procedures are time-consuming and prone to human error. YolovV5 is a model that is prominent for its precision, versatility and availability of pre-trained weights, making it a popular choice among both novices and professionals. YoloV5 allows to employ it in different domains by adjusting the model architecture, such as adding more layers, changing the head or backbone, or training it on a new dataset by changing the hyperparameters. YOLOv5 is tested to tackle a array of object identification problems in fields such as agriculture, medicine, and transportation. Our study explores variant of YOLOv5 e.g., YOLOv5n, YOLOv5s, YOLOv5m, and YOLOv5l on small set training dataset evenly distributed among all the subtypes of WBC blood cell such as neutrophils, basophils, eosinophils, monocytes, and lymphocytes further analysing which model performs better in terms of Accuracy, recall and mAP@50 and mAP@50-95, parameters used, model size accuracy and inference speed and resource utilization e.g., CPU, network and memory. Here are some of the question that our study tires to answer: Which model variant performs best at each stage of training? This is a different comparison because it focuses on model variants, not epoch counts. Is longer training worth the computational cost?