public static ObjectDetectionResult[] Detect(dynamic detector, Size supportedSize, Image <Rgb24> image)
{
if (detector is null)
{
throw new ArgumentNullException(nameof(detector));
}
if (image is null)
{
throw new ArgumentNullException(nameof(image));
}
var input = ImageTools.YoloPreprocess(new ObjectDetectionDataset.ClrEntry {
Image = image.Clone(),
}, supportedSize);
var images = input.Image[np.newaxis, np.rest_of_the_axes].AsArray();
IDictionary <string, Tensor> prediction = detector(tf.constant(images));
_ArrayLike Get(string name) => prediction["tf_op_layer_" + name].numpy();
ndarray <float> boxs = Get(nameof(SelectedBoxesOutput.Boxes)).AsArray <float>();
ndarray <float> scores = Get(nameof(SelectedBoxesOutput.Scores)).AsArray <float>();
ndarray <long> classes = Get(nameof(SelectedBoxesOutput.Classes)).AsArray <long>();
ndarray <int> detections = Get(nameof(SelectedBoxesOutput.Detections)).AsArray <int>();
return(ObjectDetectionResult.FromCombinedNonMaxSuppressionBatch(
boxs, scores, classes, detections[0].AsScalar()));
}
public static ObjectDetectionResult[] FromCombinedNonMaxSuppressionBatch(
ndarray <float> boxes, ndarray <float> scores, ndarray <long> classes,
int detectionCount)
{
var result = new ObjectDetectionResult[detectionCount];
for (int detection = 0; detection < detectionCount; detection++)
{
result[detection] = new ObjectDetectionResult {
Class = checked ((int)classes[0, detection].AsScalar()),
Box = ToBox(boxes[0, detection].AsArray()),
Score = scores[0, detection].AsScalar(),
};
}
return(result);
}
public static ObjectDetectionResult[] DetectRaw(Model rawDetector,
Size supportedSize, int classCount,
Image <Rgb24> image,
ReadOnlySpan <int> strides, Tensor <int> anchors,
ReadOnlySpan <float> xyScale,
float scoreThreshold = 0.2f)
{
if (rawDetector is null)
{
throw new ArgumentNullException(nameof(rawDetector));
}
if (image is null)
{
throw new ArgumentNullException(nameof(image));
}
var input = ImageTools.YoloPreprocess(new ObjectDetectionDataset.ClrEntry {
Image = image.Clone(),
}, supportedSize);
var images = input.Image[np.newaxis, np.rest_of_the_axes].AsArray();
IList <Tensor> prediction = rawDetector.__call__(images);
Debug.Assert(prediction.Count == 3);
var output = new YOLOv4.Output {
SSBox = prediction[0],
MBBox = prediction[1],
LBBox = prediction[2],
};
var suppression = SelectBoxes(output, inputSize: supportedSize.Width, classCount: classCount,
strides: strides, anchors: anchors,
xyScale: xyScale,
scoreThreshold: scoreThreshold);
ndarray <float> boxs = suppression.Boxes.numpy();
ndarray <float> scores = suppression.Scores.numpy();
ndarray <long> classes = suppression.Classes.numpy();
ndarray <int> detections = suppression.Detections.numpy();
return(ObjectDetectionResult.FromCombinedNonMaxSuppressionBatch(
boxs, scores, classes, detections[0].AsScalar()));
}
