Example #1
0
        public ClassifierModel TrainModel()
        {
            var attributes = Attributes.Where(x => !x.IsClass);

            //[ClassValue][Attribute][AttributeValue]
            int[][][]     countMatrix = new int[ClassInfo.NominalOptionCount][][];
            decimal[][][] modelMatrix = new decimal[countMatrix.Length][][];
            for (int i = 0; i < countMatrix.Length; i++)
            {
                countMatrix[i] = new int[attributes.Count()][];
                modelMatrix[i] = new decimal[countMatrix[i].Length][];
                int j = 0;
                foreach (var attr in attributes)
                {
                    countMatrix[i][j] = new int[attr.NominalOptionCount];
                    modelMatrix[i][j] = new decimal[attr.NominalOptionCount];
                    j++;
                }
            }
            foreach (var row in Data)
            {
                var data = GetRowData(row, Attributes);
                var classValue = data.ElementAt(ClassIndex);
                int i = 0, j = 0;
                foreach (var o in data)
                {
                    if (i != ClassIndex)
                    {
                        countMatrix[classValue][j][o]++;
                        j++;
                    }
                    i++;
                }
            }
            for (int i = 0; i < countMatrix.Length; i++)
            {
                var countBase = (decimal)countMatrix[i][0].Sum();
                if (countBase != 0)
                {
                    for (int j = 0; j < countMatrix[i].Length; j++)
                    {
                        for (int k = 0; k < countMatrix[i][j].Length; k++)
                        {
                            modelMatrix[i][j][k] = countMatrix[i][j][k] / countBase;
                        }
                    }
                }
            }
            _TrainedModel = new ClassifierModel(ClassInfo, attributes, modelMatrix);
            return(_TrainedModel);
        }
Example #2
0
        public DamerauLevensteinClassifier(ClassifierModel model)
        {
            if (model == null)
            {
                throw new ArgumentNullException(nameof(model));
            }

            MAX = model.MaxChanges;
            SHORTSYNNONYMACCEPTANCETHRESHOLD = model.ShortSynnonymAcceptanceThreshold;
            STRINGLENGTHTHRESHOLD            = model.StringLengthThreshold;
            BREAKONCOMPLETEMATCH             = model.BreakOnCompleteMatch;
            FULLSTRINGMATCHPROPORTION        = model.FullStringMatchProportion;
            EACHWORDMATCHPROPORTION          = model.EachWordMatchProportion;
            MAXITERATIVEGROUPVALUE           = model.MaxIterativeWordGroupValue;
            dLmetric = new DamerauLevensteinMetric(new DamerauLevensteinMetricModel {
                MaxLength = MAX
            });
        }
Example #3
0
        private void InitialiseClassifiers()
        {
            if (!IsSupported)
            {
                return;
            }

            var modelDir = MLUtils.ModelFolder;

            if (modelDir != null)
            {
                var haarcascades = modelDir.GetFiles("haarcascade*.xml", SearchOption.AllDirectories).ToList();

                foreach (var modelFile in haarcascades)
                {
                    try
                    {
                        var tag = modelFile.Directory?.Name.Transform(To.SentenceCase);

                        if (string.IsNullOrEmpty(tag))
                        {
                            tag = "Object";
                        }

                        var classifier = new CascadeClassifier(modelFile.FullName);
                        var model      = new ClassifierModel(modelFile.Name, tag, classifier);

                        Logging.Log($"Initialised EMGU classifier with {model.ClassifierFile} for tag '{tag}'");
                        classifiers.Add(model);
                    }
                    catch (Exception ex)
                    {
                        Logging.LogError($"Unable to initialize Emgu face detection: {ex}");
                    }
                }
            }
        }