[QUESTION] Running Hydra on ROS 2 bag - no results in RViz

[anuvabanwasi]

Hi, I've been trying to run Hydra on a sample ROS bag. I followed all the steps listed here and was able to setup the dev docker environment and launch/run Hydra via ros2 launch hydra_ros uhumans2.launch.yaml in one terminal. I converted the ROS bag from ROS1 to ROS2 and then ran ros2 bag play /root/data/path/to/rosbag --clock --qos-profile-overrides-path ~/.tf_overrides.yaml.

I can see the robot exploring the environment in RViz but it doesn't show any scene graph.

Added some print statements, and it looks like the data is not being received:

Since it's set up to use ground truth semantics, it expects one of these: - set_remap: {from: hydra/input/left_cam/semantic/image_raw, to: $(if $(var use_gt_semantics) /tesse/seg_cam/converted/image_raw /tesse/left_cam/semantic/image_raw)} but the topic doesn't exist in the rosbag:

root@lenny:~/hydra_ws# ros2 bag info src/hydra_ros/docker/uHumans2_office_s1_00h_ros2/

Files:             uHumans2_office_s1_00h_ros2.db3
Bag size:          40.5 GiB
Storage id:        sqlite3
ROS Distro:        rosbags
Duration:          506.149900000s
Start:             Jan  1 1970 00:00:11.404999999 (11.404999999)
End:               Jan  1 1970 00:08:37.554899999 (517.554899999)
Messages:          742621
Topic information: Topic: /clock | Type: rosgraph_msgs/msg/Clock | Count: 10151 | Serialization Format: cdr
                   Topic: /tesse/depth_cam/camera_info | Type: sensor_msgs/msg/CameraInfo | Count: 8307 | Serialization Format: cdr
                   Topic: /tesse/depth_cam/mono/image_raw | Type: sensor_msgs/msg/Image | Count: 8307 | Serialization Format: cdr
                   Topic: /tesse/front_lidar/scan | Type: sensor_msgs/msg/LaserScan | Count: 16962 | Serialization Format: cdr
                   Topic: /tesse/imu/clean/imu | Type: sensor_msgs/msg/Imu | Count: 101213 | Serialization Format: cdr
                   Topic: /tesse/imu/noisy/biases/accel | Type: geometry_msgs/msg/Vector3Stamped | Count: 101150 | Serialization Format: cdr
                   Topic: /tesse/imu/noisy/biases/gyro | Type: geometry_msgs/msg/Vector3Stamped | Count: 101150 | Serialization Format: cdr
                   Topic: /tesse/imu/noisy/imu | Type: sensor_msgs/msg/Imu | Count: 101212 | Serialization Format: cdr
                   Topic: /tesse/left_cam/camera_info | Type: sensor_msgs/msg/CameraInfo | Count: 8307 | Serialization Format: cdr
                   Topic: /tesse/left_cam/mono/image_raw | Type: sensor_msgs/msg/Image | Count: 8307 | Serialization Format: cdr
                   Topic: /tesse/left_cam/rgb/image_raw | Type: sensor_msgs/msg/Image | Count: 8307 | Serialization Format: cdr
                   Topic: /tesse/odom | Type: nav_msgs/msg/Odometry | Count: 101213 | Serialization Format: cdr
                   Topic: /tesse/rear_lidar/scan | Type: sensor_msgs/msg/LaserScan | Count: 16962 | Serialization Format: cdr
                   Topic: /tesse/right_cam/camera_info | Type: sensor_msgs/msg/CameraInfo | Count: 8307 | Serialization Format: cdr
                   Topic: /tesse/right_cam/mono/image_raw | Type: sensor_msgs/msg/Image | Count: 8307 | Serialization Format: cdr
                   Topic: /tesse/right_cam/rgb/image_raw | Type: sensor_msgs/msg/Image | Count: 8307 | Serialization Format: cdr
                   Topic: /tesse/seg_cam/camera_info | Type: sensor_msgs/msg/CameraInfo | Count: 8307 | Serialization Format: cdr
                   Topic: /tesse/seg_cam/rgb/image_raw | Type: sensor_msgs/msg/Image | Count: 8307 | Serialization Format: cdr
                   Topic: /tf | Type: tf2_msgs/msg/TFMessage | Count: 109537 | Serialization Format: cdr
                   Topic: /tf_static | Type: tf2_msgs/msg/TFMessage | Count: 1 | Serialization Format: cdr

There is a semantic topic for gt_semantic in the ROS 1 bag. But for some reason it’s just in rgb and it’s missing a step to convert them from rgb to label. Any advice on how to address this issue?

For context, I am using the uHumans2_office_s1_00h.bag and converted to ROS2 via rosbags-convert --src path/to/office.bag --dst path/to/office.

[anuvabanwasi]

Hi we were able to fix this issue by cloning the semantic_inference package and its dependencies via vcs import. We also had to change line 4 of uhumans2.launch.yaml to false in order to make it work:

• arg: {name: use_gt_semantics, default: 'false'}

For some reason, when running it with true it didn't work. Any idea for why this may be the case? Thanks!