BIOSKOP: AI based Microscope Design for Biological Survey and Object Analysis

Bioskop” is an advanced smart microscope that leverages artificial intelligence models, featuring an adjustable zoom capability of up to 1600x. During the development process, three AI models were trained using YOLOv11 for over 100 epochs and, these models can detect parasites, cells, and other microorganisms effectively, providing users with precise information about their locations and quantities. At its core, Bioskop utilizes Raspberry Pi 5 as the control unit, providing a robust environment for the application and its models. Researchers and users can easily train their AI models to detect microorganisms not encompassed by the three pre-trained models provided and, with the help of UI researchers, can engage in this process without needing extensive knowledge of AI, statistics, or software development. To facilitate automatic detection, users place the object on the microscope stage, capturing images of the object from multiple angles. Before these images can be employed for training an AI model, they must be labeled. Users can easily label the object by clicking on two corners. Once the labeling is complete, the training process begins. The microscope organizes the labeled images into “training” and “verification” categories. The dataset is segmented and trained over 50 epochs, with updates presented to the user as “model is being trained.” This operation runs in the background, allowing the user to focus on the task. Upon completion, users can thoroughly analyze the microorganisms they wish to detect using the “Bioskop”. We believe that this research will shed light new ideas on similar projects.