High-Resolution Colony Images of Clinically Isolated Bacteria for Automated Detection and Deep Learning.

Area of Science:

• Microbiology
• Computer Vision
• Data Science

Background:

• Manual analysis of bacterial colonies on solid media is crucial but inefficient and subjective.
• Existing datasets for AI-driven colony analysis are often limited in size, diversity, and consistency, hindering model generalization.
• There is a need for standardized, large-scale, and diverse datasets to advance automated microbiological detection.

Purpose of the Study:

• To address the limitations of existing datasets by creating a large, normalized, and diverse collection of bacterial colony images.
• To provide a robust foundation for training and evaluating AI models for automated colony detection.
• To facilitate efficient, standardized, and traceable microbiological research.

Main Methods:

• Collected a large dataset of bacterial colony images from 19 species and 151 strains under controlled conditions (closed background, stable lighting, uniform angles).
• Systematically annotated and augmented images to enhance usability for AI model development.
• Ensured diversity by including 50 images per species to capture phenotypic variations.

Main Results:

• The dataset comprises 118,442 colony instances, representing a significant resource for AI model training.
• The images cover 19 bacterial species and 151 strains, offering substantial diversity.
• The standardized collection and annotation protocols ensure data quality and usability.

Conclusions:

• The publicly released dataset provides a strong foundation for developing and validating AI models for automated colony analysis.
• This resource can overcome the limitations of previous datasets, enabling more generalized and reliable automated detection.
• The dataset supports advancements in efficient, standardized, and traceable microbiological research through AI.