How do I set the minimum threshold for drawing a box?

Question

I have an object detection model that I use with opencv to detect my custom class.

I want to output the boxes only when the model is 95% or more confident.

Is there a way I can configure that?

(Bonus question: Can I set it so that only the object with the highest confidence is shown? Example: The cam detects two objects with 98% and 91% confidence respectively. I want it to only output the box for the 98% one.)

In case you need it, here is my inference code that uses opencv.

with detection_graph.as_default():
  with tf.Session(graph=detection_graph) as sess:
    while True:
      ret, image_np = cap.read()
      image_np_expanded = np.expand_dims(image_np, axis=0)
      image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
      boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
      scores = detection_graph.get_tensor_by_name('detection_scores:0')
      classes = detection_graph.get_tensor_by_name('detection_classes:0')
      num_detections = detection_graph.get_tensor_by_name('num_detections:0')

      (boxes, scores, classes, num_detections) = sess.run(
          [boxes, scores, classes, num_detections],
          feed_dict={image_tensor: image_np_expanded})

      vis_util.visualize_boxes_and_labels_on_image_array(
          image_np,
          np.squeeze(boxes),
          np.squeeze(classes).astype(np.int32),
          np.squeeze(scores),
          category_index,
          use_normalized_coordinates=True,
          line_thickness=6)

      cv2.imshow('object detection', cv2.resize(image_np, (800, 600)))
      if cv2.waitKey(25) & 0xFF == ord('q'):
        cv2.destroyAllWindows()
        break

Answer 1

Okay, answering my own question. It only took me 1 minute to find out on my own.

This is the function that visualizes the boxes. Its input parameters are well explained in the source code.

object_detection/utils/visualization_utils.visualize_boxes_and_labels_on_image_array(...)

For my case, I just had to set

min_score_thresh=.95 and max_boxes_to_draw=1 when calling this function.