fix(traffic_light_roi_visualizer): show unknown results correctly (#9467)

fix: show unknown results correctly

Signed-off-by: tzhong518 <sworgun@gmail.com>

Author: tzhong518

perception/autoware_traffic_light_visualization/src/traffic_light_roi_visualizer/node.cpp

@@ -201,16 +201,18 @@ void TrafficLightRoiVisualizerNode::imageRoughRoiCallback(
     // bbox drawing
     cv_ptr = cv_bridge::toCvCopy(input_image_msg, sensor_msgs::image_encodings::RGB8);
     for (auto tl_rough_roi : input_tl_rough_roi_msg->rois) {
-      // visualize rough roi
-      createRect(cv_ptr->image, tl_rough_roi, cv::Scalar(0, 255, 0));
-
+      // note: a signal will still be output even if it is undetected
+      // Its position and size will be set as 0 and the color will be set as unknown
+      // So a rough roi will always have correspond roi a correspond traffic signal
       ClassificationResult result;
       bool has_correspond_traffic_signal =
         getClassificationResult(tl_rough_roi.traffic_light_id, *input_traffic_signals_msg, result);
       tier4_perception_msgs::msg::TrafficLightRoi tl_roi;
       bool has_correspond_roi =
         getRoiFromId(tl_rough_roi.traffic_light_id, input_tl_roi_msg, tl_roi);
 
+      createRect(cv_ptr->image, tl_rough_roi, extractShapeInfo(result.label).color);
+
       if (has_correspond_roi && has_correspond_traffic_signal) {
         // has fine detection and classification results
         createRect(cv_ptr->image, tl_roi, result);

perception/autoware_traffic_light_visualization/src/traffic_light_roi_visualizer/shape_draw.cpp

@@ -29,6 +29,10 @@ void drawShape(
   cv::Mat & image, const std::vector<ShapeImgParam> & params, int size, const cv::Point & position,
   const cv::Scalar & color, float probability)
 {
+  // skip if the roi position is set as (0,0), which means it is undetected
+  if (position.x == 0 && position.y == 0) {
+    return;
+  }
   // load concatenated shape image
   const auto shape_img = loadShapeImage(params, size);
 
@@ -42,14 +46,13 @@ void drawShape(
   const int fill_rect_h = std::max(shape_img.rows, text_size.height) + 10;
 
   const cv::Point rect_position(position.x, position.y - fill_rect_h);
-
   if (
     rect_position.x < 0 || rect_position.y < 0 || rect_position.x + fill_rect_w > image.cols ||
     position.y > image.rows) {
     // TODO(KhalilSelyan): This error message may flood the terminal logs, so commented out
     // temporarily. Need to consider a better way.
 
-    // std::cerr << "Adjusted position is out of image bounds." << std::endl;
+    std::cerr << "Adjusted position is out of image bounds." << std::endl;
     return;
   }
   cv::Rect rectangle(rect_position.x, rect_position.y, fill_rect_w, fill_rect_h);