MAOMaps is a dataset for evaluation of Visual SLAM, RGB-D SLAM and Map Merging algorithms. It contains 40 samples with RGB and depth images, and ground truth trajectories and maps. These 40 samples are joined into 20 pairs of overlapping maps for map merging methods evaluation. The samples were collected using Matterport3D dataset and Habitat simulator.

The dataset is available here.

To use this dataset in your research, please cite the paper: TBA.

The dataset contains 20 pairs of trajectories. Each pair is stored in its own subdirectory.

Each sample contains:

• first.bag, second.bag - raw data in Rosbag format for first and second trajectory of the pair. Rosbags contain RGB images (in topic /habitat/rgb/image), depth maps (in topic /habitat/depth/image), camera info (in topic /habitat/rgb/camera_info), and poses (in topic /true_pose).

• gt_points_first.pcd, gt_points_second.pcd - ground truth maps stored as point clouds, for first and second trajectory of the pair.

• gt_points_first.txt, gt_points_second.txt - ground truth maps coordinates in text format, for first and second trajectory of the pair.

• gt_points_merged.pcd - merged ground truth map in Pointcloud format.

• gt_points_merged.txt - merged ground truth map points in text format.

• gt_colors_first.txt, gt_colors_second.txt - colors of points of first and second ground truth map in RGB format.

• gt_colors_merged.txt - colors of points of merged ground truth map in RGB format.

• gt_poses_first.txt, gt_poses_second.txt - ground truth trajectories stores as a set of 6D poses.

• start_pose_first.txt, start_pose_second.txt - ground truth start positions of first and second trajectory.

This toolbox contains scripts for VSLAM and Map Merging algorithms evaluation (stored in evaluate folder) and scripts for dataset expanding (stored in collect_data folder).

To use this part of toolbox, clone slam_comparison repo into your ROS workspace and build it:

cd ~/catkin_ws/src
git clone https://github.com/CnnDepth/slam_comparison
cd ..; catkin_make;

Usage example: estimate RTAB-MAP quality

1. Open terminal, and run Ros master node:

roscore

2. Open another terminal, and launch the SLAM:

roslaunch rtabmap_ros rtabmap.launch

3. In third terminal, enter the dataset directory, and play some rosbag of the dataset, e.g.

rosbag play sample1/first.bag

4. Create directory to store SLAM pointclouds:

mkdir sample1/eval_rtabmap

5. In fourth terminal, run pointcloud saver:

rosrun pointcloud_processing pointcloud_subscriber /octomap_occupied_space ./sample1/eval_rtabmap/slam_points.txt ./sample1/eval_rtabmap/slam_colors.txt

6. After rosbag stops, and SLAM finishes building map, stop pointcloud saver node in fourth terminal using Ctrl-C command

7. Copy groundtruth map and trajectory into SLAM evaluation directory:

cp sample1/gt_points_first.txt sample1/eval_rtabmap/gt_points.txt
cp sample1/gt_poses_first.txt sample1/eval_rtabmap

8. Transform SLAM pointcloud into Habitat coordinate system:

python evaluate/pointcloud_transformer.py ./sample1/eval_rtabmap/slam_points.txt ./sample1/start_pose_first.txt -0.45
mv sample1/eval_rtabmap/slam_points_transformed.txt sample1/eval_rtabmap/slam_points.txt

9. Compute the absolute mapping error of the SLAM:

rosrun pointcloud_processing octomap_processing ./sample1/eval_rtabmap abs nearest ./sample1/eval_rtabmap/results.txt

To use this part of toolbox, clone habitat_ros repo into your ROS workspace and build it:

cd ~/catkin_ws/src
git clone https://github.com/CnnDepth/habitat_ros
cd ..; catkin_make;

Usage example: create 21st sample of the dataset

1. Open terminal, and run Ros master node:

roscore

2. Open another terminal, and run Habitat keyboard agent:

roslaunch habitat_ros keyboard_agent.launch

3. Create directory sample21 for new sample:

mkdir sample21

4. In third terminal, record Rosbag from Habitat:

rosbag record /habitat/rgb/image /habitat/depth/image /habitat/rgb/camera_info /true_pose -O sample21/first.bag

5. Move agent through virtual environment using arrow keys on keyboard

6. After end of movement, stop rosbag recording in third terminal using Ctrl-C command

7. Play recorded rosbag:

rosbag play sample21/first.bag

8. In fourth terminal, create ground-truth map of the rosbag:

python collect_data/gt_map_creator.py ./sample21/gt_points_first.txt ./sample21/gt_colors_first.txt 0.0

9. In fifth terminal, create ground-truth trajectory of the rosbag:

python collect_data/gt_path_writer.py ./sample21/gt_poses_first.txt

10. After rosbag ends, interrupt ROS loops in fourth and fifth terminal using Ctrl-C command.