In the pharmaceutical industry, image analysis plays a crucial role in microbial identification. Traditional parametric algorithms, while effective for tracking microbial growth over time, struggle with differentiating between bacterial and fungal species due to their rigid predefined rules. Machine learning, particularly deep learning, offers a powerful alternative by learning complex patterns from large datasets, enabling more accurate and adaptable classification. The goal here is to explore the limitations of classical algorithms, the advantages of AI-driven approaches, and the methodology for building a robust training dataset to enhance model performance. A case study on automated mold identification on petri dishes will illustrate these concepts in a real-world application.
Lisa Mallam (Thu,) studied this question.