Description

This repo contains code allowing to convert the 2D output of darknet into 3D boxes.

Dependencies

numpy
sensor_msgs
darknet_ros_msgs
sara_msgs
cv2
matplotlib
scipy

Launchfile

wm_darknet_ros_3d_wrapper.launch
- Start the 3D conversion stack.

Parameters

darknet_ros_3d_wrapper_node/depth_topic
- Select the topic from witch to get the depth images
darknet_ros_3d_wrapper_node/bounding_boxes_topic
- Select the topic from witch to get the bounding boxes
darknet_ros_3d_wrapper_node/synchroniser_buffer
- Set the buffer size for the topic synchronisation argorythm.
darknet_ros_3d_wrapper_node/synchroniser_time_tolerance
- Set the time tolerance for the topic synchronisation argorythm.
darknet_ros_3d_wrapper_node/camera_fov_width
- Set the angular width of the camera
darknet_ros_3d_wrapper_node/camera_fov_height
- Set the angular height of the camera
darknet_ros_3d_wrapper_node/histogramPrecision
- Set the number of divisions per meter used for the histogram selector
darknet_ros_3d_wrapper_node/display_gui
- Render a plot of the current segmentation. (for debug purpose only)

Nodes

bounding_box_stamp_corrector

Adjust the darknet output to have the desired timestamp.

Topics provided

Subscriber
- /darknet_ros/bounding_boxes (darknet_ros_msgs/BoundingBoxes)
  - List of bounding boxes from darknet
Publisher
- /darknet_ros/bounding_boxes_correct_stamp (darknet_ros_msgs/BoundingBoxes)
  - List of bounding boxes with adjusted timestamp

darknet_ros_3d_wrapper_node

Perform the 3D convertions of the bounding boxes.

Topics provided

Subscriber
- /head_xtion/depth/image_raw (sensor_msgs/Image)
  - Depth image from the same camera used for darknet
- /darknet_ros/bounding_boxes_correct_stamp (darknet_ros_msgs/BoundingBoxes)
  - List of bounding boxes from darknet
Publisher
- /darknet_ros_3d_wrapper_node/boxes3d (sara_msgs/BoundingBoxes3D)
  - List of 3d bounding boxes
- /darknet_ros_3d_wrapper_node/boxes3d_markers (visualization_msgs/MarkerArray)
  - List of markers for rviz