| | | 1 | | using NUnit.Framework; |
| | | 2 | | using System.Collections; |
| | | 3 | | using System.Collections.Generic; |
| | | 4 | | using System.Linq; |
| | | 5 | | using UnityEngine; |
| | | 6 | | using UnityEngine.TestTools; |
| | | 7 | | |
| | | 8 | | public class KMeansClustererTest { |
| | | 9 | | public static readonly List<GameObject> Objects = new List<GameObject> { |
| | | 10 | | GenerateObject(new Vector3(0, 0, 0)), |
| | | 11 | | GenerateObject(new Vector3(0, 1, 0)), |
| | | 12 | | GenerateObject(new Vector3(0, 1.5f, 0)), |
| | | 13 | | GenerateObject(new Vector3(0, 2.5f, 0)), |
| | | 14 | | }; |
| | | 15 | | |
| | | 16 | | public static GameObject GenerateObject(in Vector3 position) { |
| | | 17 | | GameObject obj = new GameObject(); |
| | | 18 | | obj.transform.position = position; |
| | | 19 | | return obj; |
| | | 20 | | } |
| | | 21 | | |
| | | 22 | | [Test] |
| | | 23 | | public void TestSingleCluster() { |
| | | 24 | | KMeansClusterer clusterer = new KMeansClusterer(Objects, k: 1); |
| | | 25 | | clusterer.Cluster(); |
| | | 26 | | Cluster cluster = clusterer.Clusters[0]; |
| | | 27 | | Assert.AreEqual(Objects.Count, cluster.Size()); |
| | | 28 | | Assert.AreEqual(new Vector3(0, 1.25f, 0), cluster.Coordinates); |
| | | 29 | | Assert.AreEqual(new Vector3(0, 1.25f, 0), cluster.Centroid()); |
| | | 30 | | } |
| | | 31 | | |
| | | 32 | | // Test to reveal improper clearing of cluster memberships. |
| | | 33 | | [Test] |
| | | 34 | | public void TestTwoDistinctClustersWithResetNeeded() { |
| | | 35 | | // Group A: points near (0, 0, 0). |
| | | 36 | | var groupA = new List<GameObject> { |
| | | 37 | | GenerateObject(new Vector3(0, 0, 0)), |
| | | 38 | | GenerateObject(new Vector3(1, 0, 0)), |
| | | 39 | | GenerateObject(new Vector3(0, 1, 0)), |
| | | 40 | | GenerateObject(new Vector3(1, 1, 0)), |
| | | 41 | | }; |
| | | 42 | | |
| | | 43 | | // Group B: points near (10, 10, 10). |
| | | 44 | | var groupB = new List<GameObject> { |
| | | 45 | | GenerateObject(new Vector3(10, 10, 10)), |
| | | 46 | | GenerateObject(new Vector3(11, 10, 10)), |
| | | 47 | | GenerateObject(new Vector3(10, 11, 10)), |
| | | 48 | | GenerateObject(new Vector3(11, 11, 10)), |
| | | 49 | | }; |
| | | 50 | | |
| | | 51 | | // Combine them. |
| | | 52 | | var objects = new List<GameObject>(); |
| | | 53 | | objects.AddRange(groupA); |
| | | 54 | | objects.AddRange(groupB); |
| | | 55 | | |
| | | 56 | | // Create clusterer with k = 2. |
| | | 57 | | KMeansClusterer clusterer = new KMeansClusterer(objects, k: 2); |
| | | 58 | | clusterer.Cluster(); |
| | | 59 | | |
| | | 60 | | // We expect exactly 2 clusters. |
| | | 61 | | Assert.AreEqual(2, clusterer.Clusters.Count); |
| | | 62 | | |
| | | 63 | | // Retrieve the clusters. |
| | | 64 | | Cluster c0 = clusterer.Clusters[0]; |
| | | 65 | | Cluster c1 = clusterer.Clusters[1]; |
| | | 66 | | |
| | | 67 | | // Because the clusters are well-separated, each cluster should contain all points from one |
| | | 68 | | // group or the other, not a mixture. Check via centroids. |
| | | 69 | | var centroid0 = c0.Centroid(); |
| | | 70 | | var centroid1 = c1.Centroid(); |
| | | 71 | | |
| | | 72 | | // One centroid should be near (0.5, 0.5, 0), the other near (10.5, 10.5, 10). |
| | | 73 | | var expectedCentroid0 = new Vector3(0.5f, 0.5f, 0); |
| | | 74 | | var expectedCentroid1 = new Vector3(10.5f, 10.5f, 10); |
| | | 75 | | bool correctPlacement = (centroid0 == expectedCentroid0 && centroid1 == expectedCentroid1) || |
| | | 76 | | (centroid0 == expectedCentroid1 && centroid1 == expectedCentroid0); |
| | | 77 | | Assert.IsTrue( |
| | | 78 | | correctPlacement, |
| | | 79 | | "Centroids not close to the expected group centers. Possible leftover membership from a previous iteration if cl |
| | | 80 | | |
| | | 81 | | // Additionally, we can count membership to confirm that each cluster got exactly four points |
| | | 82 | | // for a more direct check. |
| | | 83 | | int cluster0Count = c0.Size(); |
| | | 84 | | int cluster1Count = c1.Size(); |
| | | 85 | | Assert.AreEqual(8, cluster0Count + cluster1Count, |
| | | 86 | | "Total membership across clusters does not match the total number of objects."); |
| | | 87 | | |
| | | 88 | | // Even if the clusters swapped roles, each cluster should have 4 points if membership was |
| | | 89 | | // properly reset and re-assigned. |
| | | 90 | | bool clusterCountsValid = cluster0Count == 4 && cluster1Count == 4; |
| | | 91 | | Assert.IsTrue(clusterCountsValid, |
| | | 92 | | $"Cluster sizes not as expected. c0={cluster0Count}, c1={cluster1Count}."); |
| | | 93 | | } |
| | | 94 | | } |
| | | 95 | | |
| | | 96 | | public class ConstrainedKMeansClustererTest { |
| | 0 | 97 | | public static readonly List<GameObject> Objects = new List<GameObject> { |
| | | 98 | | GenerateObject(new Vector3(0, 0, 0)), |
| | | 99 | | GenerateObject(new Vector3(0, 1, 0)), |
| | | 100 | | GenerateObject(new Vector3(0, 1.5f, 0)), |
| | | 101 | | GenerateObject(new Vector3(0, 2.5f, 0)), |
| | | 102 | | }; |
| | | 103 | | |
| | 0 | 104 | | public static GameObject GenerateObject(in Vector3 position) { |
| | 0 | 105 | | GameObject obj = new GameObject(); |
| | 0 | 106 | | obj.transform.position = position; |
| | 0 | 107 | | return obj; |
| | 0 | 108 | | } |
| | | 109 | | |
| | | 110 | | [Test] |
| | 0 | 111 | | public void TestSingleCluster() { |
| | 0 | 112 | | ConstrainedKMeansClusterer clusterer = |
| | | 113 | | new ConstrainedKMeansClusterer(Objects, maxSize: Objects.Count, maxRadius: Mathf.Infinity); |
| | 0 | 114 | | clusterer.Cluster(); |
| | 0 | 115 | | Assert.AreEqual(1, clusterer.Clusters.Count); |
| | 0 | 116 | | Cluster cluster = clusterer.Clusters[0]; |
| | 0 | 117 | | Assert.AreEqual(Objects.Count, cluster.Size()); |
| | 0 | 118 | | Assert.AreEqual(new Vector3(0, 1.25f, 0), cluster.Centroid()); |
| | 0 | 119 | | } |
| | | 120 | | |
| | | 121 | | [Test] |
| | 0 | 122 | | public void TestMaxSizeOne() { |
| | 0 | 123 | | ConstrainedKMeansClusterer clusterer = |
| | | 124 | | new ConstrainedKMeansClusterer(Objects, maxSize: 1, maxRadius: Mathf.Infinity); |
| | 0 | 125 | | clusterer.Cluster(); |
| | 0 | 126 | | Assert.AreEqual(Objects.Count, clusterer.Clusters.Count); |
| | 0 | 127 | | foreach (var cluster in clusterer.Clusters) { |
| | 0 | 128 | | Assert.AreEqual(1, cluster.Size()); |
| | 0 | 129 | | } |
| | 0 | 130 | | } |
| | | 131 | | |
| | | 132 | | [Test] |
| | 0 | 133 | | public void TestZeroRadius() { |
| | 0 | 134 | | ConstrainedKMeansClusterer clusterer = |
| | | 135 | | new ConstrainedKMeansClusterer(Objects, maxSize: Objects.Count, maxRadius: 0); |
| | 0 | 136 | | clusterer.Cluster(); |
| | 0 | 137 | | Assert.AreEqual(Objects.Count, clusterer.Clusters.Count); |
| | 0 | 138 | | foreach (var cluster in clusterer.Clusters) { |
| | 0 | 139 | | Assert.AreEqual(1, cluster.Size()); |
| | 0 | 140 | | } |
| | 0 | 141 | | } |
| | | 142 | | |
| | | 143 | | [Test] |
| | 0 | 144 | | public void TestSmallRadius() { |
| | 0 | 145 | | ConstrainedKMeansClusterer clusterer = |
| | | 146 | | new ConstrainedKMeansClusterer(Objects, maxSize: Objects.Count, maxRadius: 1); |
| | 0 | 147 | | clusterer.Cluster(); |
| | 0 | 148 | | Assert.AreEqual(2, clusterer.Clusters.Count); |
| | 0 | 149 | | } |
| | | 150 | | } |