SegmentationMetrics<T>

Helpers & Utilities · AiDotNet.NeuralNetworks.Metrics

Provides metrics for evaluating segmentation and classification tasks.

For Beginners

When a model predicts labels for different parts of a 3D shape (like “leg”, “seat”, “back” for a chair), these metrics tell us how accurate those predictions are compared to the ground truth labels.

How It Works

These metrics are essential for evaluating the quality of segmentation models on 3D data like point cloud segmentation, mesh segmentation, and voxel classification.

Methods

Method Summary
double Accuracy(int[] predictions, int[] groundTruth, int ignoreIndex) Computes overall accuracy for classification/segmentation.
double MeanF1Score(int[] predictions, int[] groundTruth, int numClasses, int ignoreIndex) Computes the mean F1 score across all classes.
double MeanIoU(int[] predictions, int[] groundTruth, int numClasses, int ignoreIndex) Computes Mean Intersection over Union (mIoU) for segmentation.
double[] F1Score(int[] predictions, int[] groundTruth, int numClasses, int ignoreIndex) Computes F1 score for each class.
double[] PerClassAccuracy(int[] predictions, int[] groundTruth, int numClasses, int ignoreIndex) Computes per-class accuracy.
double[] PerClassIoU(int[] predictions, int[] groundTruth, int numClasses, int ignoreIndex) Computes per-class IoU for segmentation.
double[] Precision(int[] predictions, int[] groundTruth, int numClasses, int ignoreIndex) Computes precision for each class.
double[] Recall(int[] predictions, int[] groundTruth, int numClasses, int ignoreIndex) Computes recall for each class.
int[] ConfusionMatrix(int[] predictions, int[] groundTruth, int numClasses, int ignoreIndex) Computes the confusion matrix.