data

This is not an officially supported Google product.

JRDB for Prediction Dataset Setup

Get the JRDB Dataset

1. Go to https://jrdb.erc.monash.edu/#downloads
2. Create a User or login.
3. Download and extract JRDB 2022 Full Train Dataset to <data_path>/train_dataset.
4. Download and extract JRDB 2022 Full Test Dataset to <data_path>/test_dataset.
5. Download and extract Train Detections from the JRDB 2019 section to <data_path>/detections.

Get the Leaderboard Test Set Tracks

For the JRDB Challenge Dataset

Download and extract this leaderboard 3D tracking result to <data_path>/test_dataset/labels/PiFeNet/. Such that you have <data_path>/test_dataset/labels/PiFeNet/00XX.txt.

For the Orginal Dataset used in the Paper

Download and extract this leaderboard 3D tracking result to <data_path>/test_dataset/labels/ss3d_mot/. Such that you have <data_path>/test_dataset/labels/ss3d_mot/00XX.txt. This was the best available leaderboard tracker at the time the method was developed.

Get the Robot Odometry

Download the compressed Odometry data file here.

Extract the files and move them to <data_path>/processed/ such that you have <data_path>/processed/odoemtry/train, <data_path>/processed/odoemtry/test.

Alternatively you can extract the robot odometry from the raw rosbags yourself via extract_robot_odometry_from_rosbag.py.

Get the Preprocessed Keypoints

Download the compressed Keypoints data file here.

Extract the files and move them to <data_path>/processed/ such that you have <data_path>/processed/labels/labels_3d_keypoints/train/, <data_path>/processed/labels/labels_3d_keypoints/test/.

Create Real-World Tracks for Train Data

Run

python jrdb_train_detections_to_tracks.py --input_path=<data_path>

Dataset Folder

You should end up with a dataset folder of the following structure

- <data_path>
  - train_dataset
    - calibration
    - detections
    - images
    - labels
    - pointclouds
  - test_dataset
    - calibration
    - images
    - labels
    - pointclouds
  - processed
    - labels
      - labels_3d_keypoints
        - train
        - test
      - labels_detections_3d
    - odoemtry
      - train
      - test

Generate the Tensorflow Dataset

For the JRDB Challenge Dataset

python jrdb_preprocess_train.py --input_path=<data_path> --output_path=<output_path> --max_distance_to_robot=50.0

python jrdb_preprocess_test.py --input_path=<data_path> --output_path=<output_path> --max_distance_to_robot=50.0 --tracking_method=PiFeNet --tracking_confidence_threshold=0.01

Please note that this can take multiple hours due to the processing of the scene's pointclouds. If you do not need the pointclouds you can speed up the processing by passing --process_pointclouds=False for both.

For the Orginal Dataset used in the Paper

python jrdb_preprocess_train.py --input_path=<data_path> --output_path=<output_path> --max_distance_to_robot=15.0

python jrdb_preprocess_test.py --input_path=<data_path> --output_path=<output_path> --max_distance_to_robot=15.0 --tracking_method=ss3d_mot

Please note that this can take multiple hours due to the processing of the scene's pointclouds. If you do not need the pointclouds you can speed up the processing by passing --process_pointclouds=False for both.