Assessing Foundation Models for Mold Colony Detection with Limited Training Data

2025-10-01

AI Summary

Mold colony detection for indoor air quality assessment typically requires large-scale annotated datasets and extensive training time. Method: This study investigates the few-shot transfer capability of vision foundation models under extremely limited labeling budgets. Leveraging a custom-built dataset of 5,000 colony images—each annotated with bounding boxes and pixel-level masks—we comparatively evaluate MaskDINO and YOLOv9. Contribution/Results: MaskDINO achieves performance on par with fully trained YOLOv9 using only 150 fine-tuning images; remarkably, it sustains ~70% reliable detection accuracy even with as few as 25 annotated samples. To our knowledge, this is the first work demonstrating that general-purpose vision foundation models exhibit high data efficiency and strong generalization in fine-grained microbial image detection. The approach substantially reduces annotation costs and establishes a novel paradigm for automated microbial analysis in resource-constrained settings.

Technology Category

• Computer Vision: Other Foundations of Computer Vision
• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Security and Privacy: Data transparency and provenance

Abstract

The process of quantifying mold colonies on Petri dish samples is of critical importance for the assessment of indoor air quality, as high colony counts can indicate potential health risks and deficiencies in ventilation systems. Conventionally the automation of such a labor-intensive process, as well as other tasks in microbiology, relies on the manual annotation of large datasets and the subsequent extensive training of models like YoloV9. To demonstrate that exhaustive annotation is not a prerequisite anymore when tackling a new vision task, we compile a representative dataset of 5000 Petri dish images annotated with bounding boxes, simulating both a traditional data collection approach as well as few-shot and low-shot scenarios with well curated subsets with instance level masks. We benchmark three vision foundation models against traditional baselines on task specific metrics, reflecting realistic real-world requirements. Notably, MaskDINO attains near-parity with an extensively trained YoloV9 model while finetuned only on 150 images, retaining competitive performance with as few as 25 images, still being reliable on $approx$ 70% of the samples. Our results show that data-efficient foundation models can match traditional approaches with only a fraction of the required data, enabling earlier development and faster iterative improvement of automated microbiological systems with a superior upper-bound performance than traditional models would achieve.

Problem

• Detecting mold colonies with limited training data using foundation models
• Reducing manual annotation requirements for microbiological image analysis
• Benchmarking data-efficient models against traditional computer vision approaches

Innovation

• Using foundation models for mold detection
• MaskDINO achieves near-parity with minimal data
• Foundation models enable data-efficient microbiological automation