private List <DnnDetectedObject> OptimizeDetections(float threshold, float nmsThreshold, bool nms, List <int> classIds, List <float> confidences, List <float> probabilities, List <Rect2d> boxes)
{
int[] indices;
if (!nms)
{
indices = Enumerable.Range(0, boxes.Count).ToArray();
}
else
{
//using non-maximum suppression to reduce overlapping low confidence box
CvDnn.NMSBoxes(boxes, confidences, threshold, nmsThreshold, out indices);
_logger.LogDebug($"NMSBoxes drop {confidences.Count - indices.Length} overlapping result.");
}
var result = new List <DnnDetectedObject>();
foreach (var i in indices)
{
var box = boxes[i];
var detection = new DnnDetectedObject()
{
Index = classIds[i],
Label = Labels[classIds[i]],
Color = Colors[classIds[i]],
Probability = probabilities[i],
BoundingBox = box
};
result.Add(detection);
}
return(result);
}
private DnnDetectedObject[] ExtractYolo2Results(IEnumerable <Mat> output, Mat image, float threshold, float nmsThreshold, bool nms = true)
{
var classIds = new List <int>();
var confidences = new List <float>();
var probabilities = new List <float>();
var boxes = new List <Rect2d>();
var w = image.Width;
var h = image.Height;
var prob = output.First();
/* YOLO2 VOC output
* 0 1 : center 2 3 : w/h
* 4 : confidence 5 ~24 : class probability */
const int prefix = 5; //skip 0~4
for (int i = 0; i < prob.Rows; i++)
{
var confidence = prob.At <float>(i, 4);
if (confidence > threshold)
{
//get classes probability
Cv2.MinMaxLoc(prob.Row(i).ColRange(prefix, prob.Cols), out _, out Point max);
var classes = max.X;
var probability = prob.At <float>(i, classes + prefix);
if (probability > threshold) //more accuracy
{
//get center and width/height
var centerX = prob.At <float>(i, 0) * w;
var centerY = prob.At <float>(i, 1) * h;
var width = prob.At <float>(i, 2) * w;
var height = prob.At <float>(i, 3) * h;
float X = Math.Max(0, centerX - (width / 2.0f));
float Y = Math.Max(0, centerY - (height / 2.0f));
//put data to list for NMSBoxes
classIds.Add(classes);
confidences.Add(confidence);
probabilities.Add(probability);
boxes.Add(new Rect2d(X, Y, width, height));
}
}
}
int[] indices;
if (!nms)
{
//using non-maximum suppression to reduce overlapping low confidence box
indices = Enumerable.Range(0, boxes.Count).ToArray();
}
else
{
CvDnn.NMSBoxes(boxes, confidences, threshold, nmsThreshold, out indices);
_logger.LogInformation($"NMSBoxes drop {confidences.Count - indices.Length} overlapping result.");
}
var result = new List <DnnDetectedObject>();
foreach (var i in indices)
{
var box = boxes[i];
var detection = new DnnDetectedObject()
{
Index = classIds[i],
Label = Labels[classIds[i]],
Color = Colors[classIds[i]],
Probability = probabilities[i],
BoundingBox = box
};
result.Add(detection);
}
return(result.ToArray());
}
private DnnDetectedObject[][] ExtractYolo3SingleResults(IEnumerable <Mat> output, Mat image, float threshold, float nmsThreshold, bool nms = true)
{
DnnDetectedObject[][] results = new DnnDetectedObject[1][];
var classIds = new List <int>();
var confidences = new List <float>();
var probabilities = new List <float>();
var boxes = new List <Rect2d>();
//for nms
classIds.Clear();
confidences.Clear();
probabilities.Clear();
boxes.Clear();
var w = image.Width;
var h = image.Height;
/*
* YOLO3 COCO trainval output
* 0 1 : center 2 3 : w/h
* 4 : confidence 5 ~ 84 : class probability
*/
const int prefix = 5; //skip 0~4
foreach (var prob in output)
{
//dimensions will be 2 for single image analysis and 3 for batch analysis
//1 input Prob Dims:2 images: 0 - 300, 1 - 85
for (var i = 0; i < prob.Size(0); i++)
{
var confidence = prob.At <float>(i, 4);
//Filter out bogus results of > 100% confidence
if (confidence > threshold && confidence <= 1.0)
{
var colRange = new Range(prefix, prob.Size(1) - 1);
var maxProbIndex = prob.FindMaxValueIndexInRange <float>(i, colRange);
if (maxProbIndex == -1)
{
continue;
}
var probability = prob.At <float>(i, maxProbIndex);
if (probability > threshold) //more accuracy, you can cancel it
{
//get center and width/height
var centerX = prob.At <float>(i, 0) * w;
var centerY = prob.At <float>(i, 1) * h;
var width = prob.At <float>(i, 2) * w;
var height = prob.At <float>(i, 3) * h;
float X = Math.Max(0, centerX - (width / 2.0f));
float Y = Math.Max(0, centerY - (height / 2.0f));
//put data to list for NMSBoxes
classIds.Add(maxProbIndex - prefix);
confidences.Add(confidence);
probabilities.Add(probability);
boxes.Add(new Rect2d(X, Y, width, height));
}
}
}
}
List <DnnDetectedObject> result = OptimizeDetections(threshold, nmsThreshold, nms, classIds, confidences, probabilities, boxes);
results[0] = result.ToArray();
return(results);
}
private IList <DnnDetectedObject[]> ExtractYoloBatchedResults(IEnumerable <Mat> output, IEnumerable <Mat> image, float threshold, float nmsThreshold, bool nms = true)
{
var inputImages = image.ToList();
DnnDetectedObject[][] results = new DnnDetectedObject[inputImages.Count][];
for (int inputIndex = 0; inputIndex < inputImages.Count; inputIndex++)
{
//for nms
_classIds.Clear();
_confidences.Clear();
_probabilities.Clear();
_boxes.Clear();
var w = inputImages[inputIndex].Width;
var h = inputImages[inputIndex].Height;
/*
* YOLO3 COCO trainval output
* 0 1 : center 2 3 : w/h
* 4 : confidence 5 ~ 84 : class probability
*/
const int prefix = 5; //skip 0~4
foreach (var prob in output)
{
//dimensions will be 2 for single image analysis and 3 for batch analysis
//2 input Prob Dims:3 images: 0 - 2, 1 - 300, 2 - 85
var probabilitiesRange = new Range(prefix, prob.Size(2) - 1);
for (var i = 0; i < prob.Size(1); i++)
{
var confidence = prob.At <float>(inputIndex, i, 4);
//Filter out bogus results of > 100% confidence
if (confidence > threshold && confidence <= 1.0)
{
var maxProbIndex = prob.FindMaxValueIndexInRange <float>(inputIndex, i, probabilitiesRange);
if (maxProbIndex == -1)
{
continue;
}
var probability = prob.At <float>(inputIndex, i, maxProbIndex);
if (probability > threshold) //more accuracy, you can cancel it
{
//get center and width/height
var centerX = prob.At <float>(inputIndex, i, 0) * w;
var centerY = prob.At <float>(inputIndex, i, 1) * h;
var width = prob.At <float>(inputIndex, i, 2) * w;
var height = prob.At <float>(inputIndex, i, 3) * h;
float X = Math.Max(0, centerX - (width / 2.0f));
float Y = Math.Max(0, centerY - (height / 2.0f));
//put data to list for NMSBoxes
_classIds.Add(maxProbIndex - prefix);
_confidences.Add(confidence);
_probabilities.Add(probability);
_boxes.Add(new Rect2d(X, Y, width, height));
}
}
}
}
results[inputIndex] = OptimizeDetections(threshold, nmsThreshold, nms).ToArray();
}
return(results);
}
