Statistics
Statistical measurements within ViDi are used to validate the trained neural network's performance. Within the deep learning paradigm, validation refers to the process of evaluating a trained neural network model against a testing data set (labeled by the user, but not used in training). For example, trying to determine the state of a model (overfit versus underfit) through the collection of statistical results of multiple neural network models against a testing data set.
The use of statistical metrics in ViDi help to qualify the following:
- Estimate future performance, e.g. estimate rates of false positives.
- Optimize tool parameters (parameter search), by finding good Training Tool Parameters or Processing Tool Parameters and setting various thresholds.
- Test the reproducibility of results.
The topics in this section will help you to understand the metrics output by the Cognex ViDi Tools:
- Score histogram
- Receiver Operating Characteristic (ROC) Curve and Area Under the Curve (AUC)
- Confusion Matrix
- Accuracy, Precision, Recall and F-Score
The performance of a deep learning-based neural network model cannot be evaluated based on either of the following:
- A "grade" of the quality of the neural network model
- A "score" that is an output of the neural network model