ros wrapper issue "[detect_grasps-1] process has died ..."

[miladmalekzadeh]

Hi, I was able to run gpd without ros using eigen_params.cfg and krylon point cloud.
Then I tried the gpd_ros. The installation went ok except some warnings regarding the libpcl version. I guess it is related to having version 1.7 in ros and 1.9 for gpd itself.
I used ros_eigen_params.cfg file for ros with the realsense camera publishing pointcloud, but I see the following error:

Created GPD ....
[ INFO] [1574350027.709595006]: Waiting for point cloud to arrive ...
[ INFO] [1574350027.904605404]: Received cloud with 208106 points.
Processing cloud with 208106 points.
Voxelized cloud: 61211
Calculating surface normals ...
num_threads: 2

[detect_grasps-1] process has died [pid 26283, exit code -11, cmd /home/milad/Hans/Git/pose-estimation/gpd/catkin_ws/devel/lib/gpd_ros/detect_grasps __name:=detect_grasps __log:=/home/milad/.ros/log/908bec9c-0c5d-11ea-8f28-24ee9a933a2d/detect_grasps-1.log].
log file: /home/milad/.ros/log/908bec9c-0c5d-11ea-8f28-24ee9a933a2d/detect_grasps-1.log*
all processes on machine have died, roslaunch will exit
shutting down processing monitor...
... shutting down processing monitor complete
done

• I am not using pcl gpu

[peterhuson]

Getting this same error.
I tried running the lenet Eigen framework:

Failed to find match for field 'rgba'.
view_points:
1.5
  0
0.5
[ INFO] [1575565058.676065877]: Received cloud with 517 points, and 517 samples
Processing cloud with 517 points.
517 samples left after workspace filtering 
Voxelized cloud: 507
Calculating surface normals ...
num_threads: 4
 runtime(computeNormals): 0.0065
camera: 0, #indices: 507, #normals: 507 
Calculated 507 surface normals in 0.0066s (mode: OpenMP).
Reversing direction of normals that do not point to at least one camera ...
 reversed 0 normals
 runtime (reverse normals): 1.37531e-05
Using 500 out of 517 available samples.
[detect_grasps-1] process has died [pid 16997, exit code -11, cmd /home/phuson/catkin_ws/devel/lib/gpd_ros/detect_grasps __name:=detect_grasps __log:=/home/phuson/.ros/log/0c966eaa-1174-11ea-b27f-0c9d92c5a6b4/detect_grasps-1.log].
log file: /home/phuson/.ros/log/0c966eaa-1174-11ea-b27f-0c9d92c5a6b4/detect_grasps-1*.log
all processes on machine have died, roslaunch will exit
shutting down processing monitor...

This is launch/ur5.launch:

<launch>
  <node name="detect_grasps" pkg="gpd_ros" type="detect_grasps" output="screen"  >
    <!-- If sequential importance sampling is used (default: false) -->
<!--
    <param name="use_importance_sampling" value="false" />
-->

    <!-- What type of point cloud is used and what ROS topic it comes from -->
    <param name="cloud_type" value="2" /> <!-- 0: PointCloud2, 1: CloudIndexed, 2: CloudSamples -->
    <param name="cloud_topic" value="/cloud_stitched" />

    <!-- (optional) The ROS topic that the samples come from (default: an empty string) -->
    <param name="samples_topic" value="" />

    <!-- Filepath to the configuration file for GPD -->
    <param name="config_file" value="/home/phuson/catkin_ws/src/gpd/cfg/ros_eigen_params.cfg" />
    <!-- <param name="config_file" value="/home/phuson/catkin_ws/src/gpd/cfg/ros_vino_params.cfg" /> -->

    <param name="rviz_topic" value="plot_grasps" />

  </node>

</launch>

I also tried running the Vino framework:

Failed to find match for field 'rgba'.
view_points:
1.5
  0
0.5
[ INFO] [1575566040.499689372]: Received cloud with 462 points, and 462 samples
Processing cloud with 462 points.
462 samples left after workspace filtering 
Voxelized cloud: 452
Calculating surface normals ...
num_threads: 6
 runtime(computeNormals): 0.0038
camera: 0, #indices: 452, #normals: 452 
Calculated 452 surface normals in 0.0039s (mode: OpenMP).
Reversing direction of normals that do not point to at least one camera ...
 reversed 0 normals
 runtime (reverse normals): 1.27731e-05
Estimating local reference frames ...
Estimated 462 frames in 0.0012s.
Finding hand poses ...
[detect_grasps-1] process has died [pid 17327, exit code -11, cmd /home/phuson/catkin_ws/devel/lib/gpd_ros/detect_grasps __name:=detect_grasps __log:=/home/phuson/.ros/log/0c966eaa-1174-11ea-b27f-0c9d92c5a6b4/detect_grasps-1.log].
log file: /home/phuson/.ros/log/0c966eaa-1174-11ea-b27f-0c9d92c5a6b4/detect_grasps-1*.log
all processes on machine have died, roslaunch will exit
shutting down processing monitor...

[FrankLin9981]

Excuse me, I have the same issue.
Is there any solution?
Thanks.

[jediofgever]

@miladmalekzadeh were you able to solve this ?

[jediofgever]

interesting ...
in ros_vino_params.cfg,
when I set weights_file to a directory it starts to work.

I change weights_file as;

weights_file = /home/atas/gpd/models/lenet/15channels/params/

[atenpas]

@jediofgever If you want to use the customary neural network implementation based on Eigen, then weights_file should be the path to a directory with weight files (like here). If you want to use OpenVINO, then weights_file should be the path to the bin file with the weights (like here).