Use custom stereo bagfiles with Isaac Visual SLAM

Hey everyone,

I’m trying to run Isaac ROS Visual SLAM on recorded bagfiles from a RealSense D455 (stereo images) but I can’t get the SLAM node to start — no images appear on the remapped topics.

What I did:

  1. Started the container and installed isaac_ros_visual_slam.
  2. Launched the SLAM node: ros2 launch isaac_ros_visual_slam isaac_ros_visual_slam.launch.py use_sim_time:=True enable_imu_fusion:=False
  3. Played the bag with remaps: ros2 bag play /workspaces/isaac_ros-dev/bagfiles/mybagfile
    --clock
    --remap /camera/camera/infra1/image_rect_raw:=/visual_slam/image_0
    --remap /camera/camera/infra2/image_rect_raw:=/visual_slam/image_1
    --remap /camera/camera/infra1/camera_info:=/visual_slam/camera_info_0
    --remap /camera/camera/infra2/camera_info:=/visual_slam/camera_info_1
    --remap /tf_static:=/tf_static_fake_topic

(I also tried remapping to /stereo_camera/left/image_rect, etc., since I wasn’t sure of the required topic names.)

My bagfile contains the following topics:

Topic information: Topic: /camera/camera/infra1/camera_info | Type: sensor_msgs/msg/CameraInfo | Count: 19595 | Serialization Format: cdr

                   Topic: /camera/camera/infra1/image_rect_raw | Type: sensor_msgs/msg/Image | Count: 19595 | Serialization Format: cdr

                   Topic: /camera/camera/infra2/camera_info | Type: sensor_msgs/msg/CameraInfo | Count: 19594 | Serialization Format: cdr

                   Topic: /camera/camera/infra2/image_rect_raw | Type: sensor_msgs/msg/Image | Count: 19592 | Serialization Format: cdr

                   Topic: /livox/points | Type: sensor_msgs/msg/PointCloud2 | Count: 6506 | Serialization Format: cdr

                   Topic: /tf_static | Type: tf2_msgs/msg/TFMessage | Count: 2 | Serialization Format: cdr

Problem: With the remaps above the Visual SLAM system does not start and no images are published to the remapped topics.

Any suggestions to get images accepted by the SLAM node, or diagnostics I should run? Thanks!