Expand description
Swarm W — evaluation metrics (synthetic always; real H1 when corpus supplied).
Synthetic and real metrics are never mixed into one score.
Structs§
Functions§
- evaluate_
real_ held_ out - Placeholder for H1 real eval: caller supplies (image, gt) pairs. Returns None if empty — never invents real metrics.
- evaluate_
synthetic - Run detector on
nsynthetic test samples; returns mean match accuracy vs GT. - mean_
best_ iou - Mean IoU of best-matching same-class pairs (diagnostic).