Method ParseYOLOv5

src/Detector/OCVDNNDetector.cpp:515–567

\brief OCVDNNDetector::ParseYOLOv5 \param crop \param detections \param tmpRegions

/// \param tmpRegions
///
void OCVDNNDetector::ParseYOLOv5(const cv::Rect& crop, std::vector<cv::Mat>& detections, regions_t& tmpRegions)
{
	int rows = detections[0].size[1];
	int dimensions = detections[0].size[2];

	// yolov5 has an output of shape (batchSize, 25200, 85) (Num classes + box[x,y,w,h] + confidence[c])
	// yolov8 has an output of shape (batchSize, 84,  8400) (Num classes + box[x,y,w,h])
	if (dimensions > rows) // Check if the shape[2] is more than shape[1] (yolov8)
	{
		rows = detections[0].size[2];
		dimensions = detections[0].size[1];

		detections[0] = detections[0].reshape(1, dimensions);
		cv::transpose(detections[0], detections[0]);
	}
	float* data = (float*)detections[0].data;

	float x_factor = crop.width / static_cast<float>(m_inWidth);
	float y_factor = crop.height / static_cast<float>(m_inHeight);

	for (int i = 0; i < rows; ++i)
	{
		float confidence = data[4];

		if (confidence >= m_confidenceThreshold)
		{
			float* classes_scores = data + 5;

			cv::Mat scores(1, static_cast<int>(m_classNames.size()), CV_32FC1, classes_scores);
			cv::Point class_id;
			double maxClassScore = 0;
			cv::minMaxLoc(scores, 0, &maxClassScore, 0, &class_id);

			if (maxClassScore > m_confidenceThreshold)
			{
				float x = data[0];
				float y = data[1];
				float w = data[2];
				float h = data[3];

				int left = cvRound((x - 0.5f * w) * x_factor);
				int top = cvRound((y - 0.5f * h) * y_factor);

				int width = cvRound(w * x_factor);
				int height = cvRound(h * y_factor);

				if (m_classesWhiteList.empty() || m_classesWhiteList.find(T2T(class_id.x)) != std::end(m_classesWhiteList))
					tmpRegions.emplace_back(cv::Rect(left + crop.x, top + crop.y, width, height), T2T(class_id.x), static_cast<float>(maxClassScore));
			}
		}
		data += dimensions;
	}
}

///
/// \brief OCVDNNDetector::ParseYOLOv8