Ground Truth
The verified, accurate data used as the reference standard for training, validating, and evaluating machine learning models.
What is Ground Truth?
Ground truth represents the correct or expected answer against which a model’s output is compared. It is often created through human annotation, direct observation, trusted records, or verified measurements. For example, in image classification, the confirmed label assigned to an image serves as ground truth when evaluating whether the model classified it correctly.
Why is Ground Truth Important?
Machine learning models need reliable reference data to measure how accurately they perform. Inaccurate or inconsistent ground truth can lead to misleading evaluation results and poor model performance. High-quality ground truth helps teams train models effectively, calculate performance metrics, and identify prediction errors.
Common use cases
Ground truth is commonly used in supervised learning, model training, model evaluation, computer vision, natural language processing, and data annotation.