TQU-SLAM Benchmark Feature-based Dataset for Building Monocular VO

Abstract

This paper introduces the TQU-SLAM benchmark dataset, which includes 160,631 RGB-D frame pairs with the goal to be used in Dell Learning (DL) training of Visual SLAM and Visual Odometry (VO) construction models. It was collected from the corridors of three interconnected buildings with a length of about 230 m. The ground-truth data of the TQU-SLAM benchmark dataset, including the 6-DOF camera pose, 3D point cloud data, intrinsic parameters, and the transformation matrix between the camera coordinate system and the real world, were prepared manually. The TQU-SLAM benchmark dataset was tested based on the PySLAM framework with traditional features, such as SHI_TOMASI, SIFT, SURF, ORB, ORB2, AKAZE, KAZE, and BRISK and features extracted from DL LIKE VGG. Experiments were also conducted on DPVO for VO estimation. The camera pose estimation results were evaluated and presented in detail, while the challenges of the TQU-SLAM benchmark dataset were analyzed.