Skip to content

Latest commit

 

History

History
131 lines (86 loc) · 5.91 KB

README.md

File metadata and controls

131 lines (86 loc) · 5.91 KB
Raw

UMAD: University of Macau Anomaly Detection Benchmark Dataset

IROS, 2024.
Dong Li, Lineng Chen, Cheng-Zhong Xu, Hui Kong†

University of Macau
†Corresponding Authors

Paper Code&Datasets Video Poster

UMAD: University of Macau Anomaly Detection Benchmark Dataset

UMAD Dataset Google Drive Link, about 10GB, and UMAD-Dataset-Usage-Guide-Doc

😊News

This work is maintaining. You can hit the STAR and WATCH to follow the updates.

  • 2024-9-5: We have released the UMAD-1.0 dataset, along with the robot system code.

  • 2024-8-27: We will update the UMAD-homo-eva dataset and the extension experiments on the UMAD-homo-eva.

  • 2024-8-22: UMAD paper sharing on arXiv~

  • 2024/6/30: UMAD has been accepted by IROS 2024! Thanks to everyone who participated in this project!

  • 2024/3/21: We have publicly released a supplementary video for the paper submission.

📝ToDo List

  • Make the project paper publicly available.
  • Open-source the UMAD dataset.
  • Open-source the UMAD-homo-eval dataset.
  • Open-source the code related to the datasets.
  • Open source robotic system code.
  • Release C++/python Adaptive Warping code.

Dataset

Dataset Overview

2

3-07-00001668-and-6-21-00003570

You can refer to the UMAD-Dataset-Usage-Guide-Doc for information on how to use the UMAD dataset and details about the ground truth mask files.

Benchmark

Anomaly Detection Benchmark

Change Detection Benchmark

System

3

You can easily collect data or deploy a system like our UMAD robot system:

# Prerequisites: [FAST_LIO](https://github.com/hku-mars/FAST_LIO) and [FAST_LIO_LOCALIZATION](https://github.com/HViktorTsoi/FAST_LIO_LOCALIZATION)

# 1. Build develop environment: Download UMAD's code, and put src/FAST_LIO_LOCALIZATION in the workspace of ROS
git clone https://github.com/IMRL/UMAD
#catkin make

# 2. Build map and record path: Put the robot in the scene, run FAST_LIO and record the waypoints
roslaunch fast_lio mapping_mid360.launch
python3 UMAD/robot_system_code/script/path_record.py # generate a path.txt file

# 3.Control the robot around the environment
# 4.Shut down the fast_lio and waypoint_record scripts.
# 5.Save the scene map output from FAST-LIO

# 6. Collect reference data: put robot back start point, run FAST_LIO_LOCALIZATION and path follow code
python3 UMAD/robot_system_code/script/path_follow.py
rosrun fast_lio_localization publish_initial_pose.py 0 0 0 0 0 0
roslaunch fast_lio_localization localization_mid360.launch map:=/home/imrl/Desktop/3.Central-Avenue.pcd
rosbag record /camera/color/image_raw/compressed /localization

# Assuming a long time has passed, or you have placed some anomalous Objects in the scene.

# 7.Collect query data: put robot back start point, run FAST_LIO_LOCALIZATION and path follow code like 6
python3 UMAD/robot_system_code/script/path_follow.py
rosrun fast_lio_localization publish_initial_pose.py 0 0 0 0 0 0
roslaunch fast_lio_localization localization_mid360.launch map:=/home/imrl/Desktop/3.Central-Avenue.pcd
rosbag record /camera/color/image_raw/compressed /localization

Acknowledgement

The authors would like to thank the following people for their contributions to data collection and data annotation for this project: @Xiangyu QIN, @Shenbo WANG, @Kaijie YIN, @Shuhao ZHAI, @Xiaonan LI, @Beibei ZHOU, and @Hongzhi CHENG.

License

Our datasets and code is released under the MIT License (see LICENSE file for details).

Citing

If you find our work useful, please consider citing:

@article{li2024umad
  author    = {Li, Dong and Chen, Lineng and Xu, Cheng-Zhong and Kong, Hui},
  title     = {UMAD: University of Macau Anomaly Detection Benchmark Dataset},
  journal   = {arXiv preprint arXiv:2408.12527},
  year      = {2024},
}

Note

You can contact Dong Li via email([email protected]) or open an issue on UMAD repo directly If you have any questions.