YOLO-based Deep Learning to Automated Bacterial Colony Counting
Jaken Whipp, Aijuan Dong
Abstract
Colony Counting is a key step in vaccine development, and manual counting is a tedious, error-prone, and labor-intensive process. In this study, we developed and evaluated multiple deep learning models for automated microbial colony counting based on the YOLO (You Only Look Once) framework. With S. aureus images from the AGAR dataset, the models achieved [email protected] between 96%-99%. Moreover, we found more complex models did not lead to much better performance. With GPUs (Graphic Processing Units) available under the Google Colab Pro, the inference time per image is about 9 milliseconds for the small YOLOv5 model. This study showed that YOLO-based deep learning models are promising in automated, real-time microbial colony counting.