dgriesse/lake_constance_obstacle_dataset
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
No description
8 commits 1 branch 0 tags 24 MiB
  • Python 100%
Find a file
Exact
Dennis Grießer 3ba1743ecd benchmark for ICRa submission
2023-09-15 15:14:14 +02:00
test benchmark for ICRa submission 2023-09-15 15:14:14 +02:00
train benchmark for ICRa submission 2023-09-15 15:14:14 +02:00
.gitattributes final test 2023-09-15 13:38:03 +02:00
read.py benchmark for ICRa submission 2023-09-15 15:14:14 +02:00
README.md benchmark for ICRa submission 2023-09-15 15:14:14 +02:00

README.md

The Lake Constance Obstacle Detection Dataset

The Lake Constance Obstacle Dataset is a dataset for object detection on inland waters. It contains stereo images, lidar point clouds, calibration data, and annotated 3d bounding boxes. There is a training dataset with 1664 files and a test dataset with 310 files. The data is saved in the hdf5 file format and can be read with for example h5py. The read.py is a script to read the data for one hdf5 file.

hdf5 file structure:

  • pointcloud
  • left_image
    • image
    • K
    • D
    • R
    • P
  • right image
    • image
    • K
    • D
    • R
    • P
  • calib_lidat_to_cam
    • R
    • t
  • calib_cam_r_to_cam_l
    • R
    • t
  • calib_cam_to_plane
    • R
    • t
  • imu_orientation
  • bounding_boxes
    • bounding_box_0
      • category
        • int
        • name
      • location
      • dimensions
      • rotation_y
      • visibility
      • occlusion
    • ...
    • bounding_box_n
      • category
        • int
        • name
      • location
      • dimensions
      • rotation_y
      • visibility
      • occlusion