Development of an Automated System for Detecting and Counting Colony-Forming Units on Agar Plates Using Computer Vision and Raspberry Pi

The manual enumeration of Colony Forming Units (CFUs) is a fundamental but labour-intensive and error-prone practice in microbiology. While commercial automated counters exist, their high cost limits accessibility. This research developed a low-cost automated CFU detection system using a Raspberry Pi and computer vision. The system integrates a camera module with controlled lighting to capture images of nutrient agar plates, which are then processed using deep learning-based analysis. A YOLOv8n model was trained and deployed on the Raspberry Pi to detect and count colonies in real time, demonstrating the feasibility of lightweight deep learning on resource-constrained hardware. Training and validation showed high precision and moderate recall, meaning the system reliably identified colonies with minimal false positives while occasionally missing some. Despite this trade-off, the system consistently detected colonies across multiple test plates, proving robustness under varying conditions. This project therefore provides a viable proof of concept for an affordable and reliable automated CFU counter. The integration of hardware and software demonstrates that cost-effective embedded systems can support practical microbiological applications. Future work should refine the model to improve recall, expand the dataset for colony diversity, and optimize the hardware setup for higher robustness and usability.