Post-process DNN object detection model predictions

detections = cv.InferBbox(delta_bbox, class_scores, conf_scores);
detections = cv.InferBbox(..., 'OptionName',optionValue, ...);

Input

• delta_bbox Blob containing relative coordinates of bounding boxes.
• class_scores Blob containing the probability values of each class.
• conf_scores Blob containing the confidence scores.

Output

• detections Struct-array which holds the final detections of the model. Each structure holds the details pertaining to a single bounding box:
  • bbox bounding box of the form [xmin, ymin, xmax, ymax]
  • class_idx class index.
  • label_name class label.
  • class_prob probability value.

Options

• Threshold threshold to filter the bounding boxes. default 0.8

This function is used with Convolutional Neural Networks for detecting objects in an image.