Dataset for evaluation of range-based people tracker classifiers in mobile robots

This dataset can be used to evaluate the performance of different approaches for detecting and tracking people by using lidar sensors. Information contained in the dataset is especially suitable to be used as test data for neural network-based classifiers.

This dataset contains 25 Rosbag files recorded in different locations with Orbi-One robot stood still. Two sorts of Rosbag files have been recorded. In 17 Rosbag files (1-17), there were people stood still in the scene. They were placed in known locations to get ground-truth data. The locations where the people were placed for each rosbag are the following:

1.bag: [1]
2.bag: [1, 2]
3.bag: [1, 2, 3]
4.bag: [2, 3, 4]
5.bag: [1, 2, 4]
6.bag: [5, 6, 7]
7.bag: [6, 7]
8.bag: [6, 7, 8]
9.bag: [11, 12, 13, 14]
10.bag: [11, 12, 13, 14, 15]
11.bag: [11, 12, 13]
12.bag: [13, 15]
13.bag: [14, 15]
14.bag: [10, 11]
15.bag: [11]
16.bag: [6]
17.bag: [6, 7]

The (x, y) positions of each point on the map are the following:

1: [1.30, 0.76]
2: [2.10, 1.56]
3: [2.90, 1.16]
4: [3.70, 0.55]
5: [6.53, 1.75]
6: [7.73, 1.16]
7: [8.93, 1.75]
8: [9.73, 0.75]
9: [14.16, 1.14]
10: [15.36, 0.14]
11: [16.56, 1.76]
12: [16.96, 0.14]
13: [17.76, 0.54]
14: [18.16, 1.54]

The remaining 8 Rosbag files (18-25) were recorded without people in the scene in order to evaluate the True Negatives rate.