DiSCount

DiSCount automates detection and quantification of Striga seeds in microscopy images to enable high-throughput assessment of seed germination and the efficacy of control measures.

Key Features:

• Computer vision and machine learning: Uses computer vision and a machine learning-based approach for automated seed detection and counting.
• Small-seed resolution: Detects Striga seeds that are less than 200 μm in size.
• Training dataset: Model development used 98 manually annotated images for training.
• Validation and testing: Performance was validated and tested against 188 manually counted images.
• Accuracy: Achieves an average error of 3.38 percentage points per image compared to manual counts.
• Throughput and performance: Processes images in approximately 3 seconds on a single CPU and ~0.1 seconds on a GPU, yielding a 100–3000-fold acceleration over manual analysis.
• Screening capability: Suited for large-scale screening of chemical compounds and biological control agents targeting Striga seed germination.

Scientific Applications:

• High-throughput screening: Enables large-scale screening of chemical compounds and biological control agents for inhibition of Striga seed germination.
• Quantitative evaluation of control measures: Provides quantitative counts to assess the impact of interventions on Striga seed germination.
• Agricultural and ecological research: Facilitates rapid generation of seed count data for studies of Striga biology, life cycle, and population responses to treatments.

Methodology:

Computer vision and machine learning model trained on 98 manually annotated images and validated/tested on 188 manually counted images.