remove score

msg/CosyObject.msg

 string label
-
 geometry_msgs/Pose pose
-
-#(Ignored for warm-start)
-float32 score

msg/CosyObjectArray.msg

 std_msgs/Header header
-
 ros_cosypose/CosyObject[] objects

scripts/pose_estimation.py

@@ -65,7 +65,6 @@ class PredictionService:
             labels = []
             poses = []
             batch_im_ids = []
-            scores = []
             for object in req.objects_prior:
                 position = [object.pose.position.x, object.pose.position.y, object.pose.position.z]
                 orientation = [object.pose.orientation.x, object.pose.orientation.y, object.pose.orientation.z, object.pose.orientation.w]
@@ -75,8 +74,7 @@ class PredictionService:
                 poses.append(poseMatrix.tolist())
                 labels.append(object.label)
                 batch_im_ids.append(0)
-                scores.append(object.score)
-            pose_estimation_prior = PandasTensorCollection(infos=pd.DataFrame(dict(label=labels,batch_im_id=batch_im_ids,score=scores)),
+            pose_estimation_prior = PandasTensorCollection(infos=pd.DataFrame(dict(label=labels,batch_im_id=batch_im_ids)),
                                                            poses=torch.tensor(poses).float().cuda())
 
         # Predict poses using cosypose
@@ -91,8 +89,7 @@ class PredictionService:
             label = pose_predictions.infos['label'][i]
             poseKDL = pm.fromMatrix(pose_predictions.poses[i])
             poseROS = Pose(position=Point(*poseKDL.p) , orientation=Quaternion(*poseKDL.M.GetQuaternion()))
-            score = pose_predictions.infos['score'][i]
-            objects_array.objects.append(CosyObject(label=label, pose=poseROS, score=score))
+            objects_array.objects.append(CosyObject(label=label, pose=poseROS))
 
         return objects_array
 

scripts/rviz_translator.py

@@ -24,9 +24,9 @@ class RVizTranslator:
                 marker.id = i
                 marker.type = 10 # Custom mesh
                 marker.action = 0 # Add/modify
-                marker.color.r = 1 - data.objects[i].score
-                marker.color.g = data.objects[i].score
-                marker.color.b = 0
+                marker.color.r = 0.8
+                marker.color.g = 0.8
+                marker.color.b = 0.8
                 marker.color.a = 0.8
                 marker.pose = data.objects[i].pose
                 marker.scale.x = .001 # mm -> m

srv/GetPrediction.srv

@@ -4,7 +4,7 @@ float64[9]  K
 # Input image
 sensor_msgs/Image image
 
-# If not empty will be used to warm start cosypose. (score field ignored)
+# If not empty will be used to warm start cosypose
 CosyObject[] objects_prior
 
 ---