BudFinder: Frame-Level Agreement Scores for Budding Event Detection in Yeast

Frame-level agreement scores between the BudFinder model and manual annotations for budding events across three yeast genetic backgrounds. The dataset, shared by Phuc Nguyen on figshare in May 2026, contains percentages of predicted events matching ground truth exactly or within tolerance windows of ±1 or ±2 frames. The model captured >78% of events within ±1 frame and >89% within ±2 frames across strains, including an oscillator and a sir2Δ mutant.

Use Cases

• Benchmarking temporal precision of cell division detection models based on frame-level agreement scores.
• Evaluating model robustness across genetic backgrounds based on performance metrics for wild-type, oscillator, and sir2Δ strains.
• Calibrating tolerance windows for event detection in time-lapse microscopy based on the ±1 and ±2 frame match criteria.

Strengths

• Performance metrics are provided for three distinct genetic backgrounds (wild-type, oscillator, sir2Δ), allowing for comparative analysis.
• Specific, quantified performance thresholds are stated: >78% agreement within ±1 frame and >89% agreement within ±2 frames.

Limitations

• The dataset is very small at 5.5 KB, indicating limited scope and likely containing only summary statistics.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for certain analyses.

Provenance

Source

Phuc Nguyen via figshare.

Collection Method

Likely generated by comparing outputs of the BudFinder model against manually annotated ground-truth data.

Freshness

Last updated 2026-05-18 17:42:39; freshness should be verified.