FScore<T>

Models & Types · AiDotNet.Metrics

F-Score metric for 3D reconstruction evaluation.

How It Works

F-Score combines precision and recall at a given distance threshold. Precision = fraction of predicted points within threshold of a ground truth point. Recall = fraction of ground truth points within threshold of a predicted point. F-Score = 2 * (Precision * Recall) / (Precision + Recall)

Usage in 3D AI:

  • 3D reconstruction quality evaluation
  • Mesh surface accuracy assessment
  • Point cloud completion evaluation

Constructors

Constructor Summary
FScore(double threshold) Initializes a new instance of the F-Score metric.

Methods

Method Summary
T Compute(Tensor<T> predicted, Tensor<T> groundTruth) Computes F-Score between predicted and ground truth point clouds.
ValueTuple<T, T> ComputePrecisionRecall(Tensor<T> predicted, Tensor<T> groundTruth) Computes precision and recall separately.