Skip to content

Model Evaluation

After running a model, RIME can evaluate its output against your gold-standard annotations.

Opening the Model Evaluation panel

View → Model Evaluation

What the panel computes

Evaluation compares model-detected events to annotated events in the corresponding lane.

Metric Description
Sensitivity True positive rate (detected FOG / annotated FOG)
Specificity True negative rate
F1 score Harmonic mean of precision and recall
Event overlap Per-episode overlap between detected and annotated events

Evaluation window

You can restrict evaluation to a specific task condition to compare model performance across walking contexts.

Exporting evaluation results

Evaluation metrics are included in the session export. See Export & Reports.