Deep Learning for Dense Colony Detection and Classification in Microbial Cultures

Since bacteria are microscopic, they need to grow and stack on solid culture media, gradually forming colonies visible to the naked eye.In addition to identifying bacterial species, calculating the number of colony-forming units (CFU) enables estimation of the bacterial count in the original sample.CFU enumeration in culture media is particularly critical for industries such as pharmaceuticals and food production.Accurate CFU counts are essential for ensuring product safety and quality, making this process a key step in manufacturing.The used AGAR dataset involves the detection and identification of countable colonies, including Staphylococcus aureus, Bacillus subtilis, Pseudomonas aeruginosa, Escherichia coli, and Candida albicans.By leveraging the imaging characteristics of bacterial colonies on different culture media, this study integrates an improved RCNN detection model and a Transformer classifier to provide more accurate judgments.