diff --git a/node_script/node.py b/node_script/node.py
index 81cb306..c89034d 100755
--- a/node_script/node.py
+++ b/node_script/node.py
@@ -2,7 +2,7 @@
 from typing import Optional
 
 import rospy
-from jsk_recognition_msgs.msg import LabelArray, VectorArray
+from jsk_recognition_msgs.msg import LabelArray, RectArray, VectorArray
 from node_config import NodeConfig
 from rospy import Publisher, Subscriber
 from sensor_msgs.msg import Image
@@ -24,6 +24,7 @@ class DeticRosNode:
     pub_segimg: Optional[Publisher]
     pub_labels: Optional[Publisher]
     pub_score: Optional[Publisher]
+    pub_rects: Optional[Publisher]
 
     # otherwise, the following publisher will be used
     pub_info: Optional[Publisher]
@@ -45,6 +46,7 @@ def __init__(self, node_config: Optional[NodeConfig] = None):
                 self.pub_segimg = rospy.Publisher('~segmentation', Image, queue_size=1)
                 self.pub_labels = rospy.Publisher('~detected_classes', LabelArray, queue_size=1)
                 self.pub_score = rospy.Publisher('~score', VectorArray, queue_size=1)
+                self.pub_rects = rospy.Publisher('~rects', RectArray, queue_size=1)
             else:
                 self.pub_info = rospy.Publisher('~segmentation_info', SegmentationInfo,
                                                 queue_size=1)
@@ -77,9 +79,11 @@ def callback_image(self, msg: Image):
             seg_img = raw_result.get_ros_segmentaion_image()
             labels = raw_result.get_label_array()
             scores = raw_result.get_score_array()
+            rects = raw_result.get_rect_array()
             self.pub_segimg.publish(seg_img)
             self.pub_labels.publish(labels)
             self.pub_score.publish(scores)
+            self.pub_rects.publish(rects)
         else:
             assert self.pub_info is not None
             seg_info = raw_result.get_segmentation_info()
diff --git a/node_script/wrapper.py b/node_script/wrapper.py
index 1f0c3ff..bddb706 100644
--- a/node_script/wrapper.py
+++ b/node_script/wrapper.py
@@ -10,7 +10,7 @@
 from cv_bridge import CvBridge
 from detectron2.utils.visualizer import VisImage
 from detic.predictor import VisualizationDemo
-from jsk_recognition_msgs.msg import Label, LabelArray, VectorArray
+from jsk_recognition_msgs.msg import Label, LabelArray, Rect, RectArray, VectorArray
 from node_config import NodeConfig
 from sensor_msgs.msg import Image
 from std_msgs.msg import Header
@@ -28,6 +28,7 @@ class InferenceRawResult:
     visualization: Optional[VisImage]
     header: Header
     detected_class_names: List[str]
+    boxes: List[List[float]]
 
     def get_ros_segmentaion_image(self) -> Image:
         seg_img = _cv_bridge.cv2_to_imgmsg(self.segmentation_raw_image, encoding="32SC1")
@@ -68,6 +69,14 @@ def get_segmentation_info(self) -> SegmentationInfo:
                                     header=self.header)
         return seg_info
 
+    def get_rect_array(self) -> RectArray:
+        rects = [Rect(x=int(box[0]),
+                      y=int(box[1]),
+                      width=int(box[2] - box[0]),
+                      height=int(box[3] - box[1])) for box in self.boxes]
+        rec_arr = RectArray(header=self.header, rects=rects)
+        return rec_arr
+
 
 class DeticWrapper:
     predictor: VisualizationDemo
@@ -122,12 +131,14 @@ def infer(self, msg: Image) -> InferenceRawResult:
         pred_masks = list(instances.pred_masks)
         scores = instances.scores.tolist()
         class_indices = instances.pred_classes.tolist()
+        boxes = list(instances.pred_boxes)
 
         if len(scores) > 0 and self.node_config.output_highest:
             best_index = np.argmax(scores)
             pred_masks = [pred_masks[best_index]]
             scores = [scores[best_index]]
             class_indices = [class_indices[best_index]]
+            boxes = [boxes[best_index]]
 
             if self.node_config.verbose:
                 rospy.loginfo("{} with highest score {}".format(self.class_names[class_indices[0]], scores[best_index]))
@@ -150,5 +161,6 @@ def infer(self, msg: Image) -> InferenceRawResult:
             scores,
             visualized_output,
             msg.header,
-            detected_classes_names)
+            detected_classes_names,
+            boxes)
         return result