Training Metrics

Audio Classification & Gesture Recognition

Metric	Description
Training Accuracy	Accuracy on training data
Validation Accuracy	Accuracy on validation/test data
Training Loss	Loss value on training data
Validation Loss	Loss value on validation/test data

Image Classification

Metric	Description
Box Loss	Accuracy of predicted bounding box locations — lower is better
Cls Loss	Accuracy of predicted object class — lower is better
DFL Loss	Distribution Focal Loss — refines bounding box predictions — lower is better
Precision	Fraction of detections that were correct — higher is better
Recall	Fraction of actual objects successfully detected — higher is better
mAP50	Mean Average Precision at IoU 0.50 — primary accuracy metric — higher is better
mAP50-95	mAP averaged across IoU 0.50–0.95 — stricter measure — higher is better