Model Evaluation for Roboflow Instant Models

You can now run Model Evaluations on Roboflow Instant models.

This release lets you:

  • Access tools like the Confusion Matrix, Vector Analysis, and view Precision and Recall metrics for Instant models

  • Compare Ground Truth vs Predictions in the image viewer

  • Runs on validation and test set images

  • Supports custom confidence thresholds for Instant (90–99%) → Use the confidence slider to explore optimal confidence

Evaluations trigger automatically after Instant training if your dataset contains images in the valid or test datasets. You can also trigger a model evaluation manually for existing models as long as you have valid or test images in your dataset.

This feature is available for all Roboflow Instant models.

Last updated

Was this helpful?