SJTU_GVI

A GNSS visual inertial dataset released as part of the tests for the M2C-GVIO paper, with raw GNSS measurements from a Ublox ZED-F9P receiver.

SJTU GVI is a GNSS visual inertial dataset for SLAM, pairing visual and inertial readings with GNSS. The README says it is part of the dataset for tests in the paper M2C-GVIO, and it lists features that differentiate it from the M2DGR dataset. The key one is that GNSS raw measurements are captured by a Ublox ZED-F9P receiver, which the authors say facilitates GNSS-SLAM research. That receiver choice is what sets the raw GNSS data apart from other collections.

The README asks users to star the project if it helps their research and to cite the related papers in academic work. Two citations are provided as BibTeX blocks. M2DGR is a multi-sensor and multi-scenario SLAM dataset for ground robots, published in IEEE Robotics and Automation Letters in 2021. M2C-GVIO is a motion manifold constraint aided GNSS visual inertial odometry paper from 2023. Both entries are formatted so researchers can copy them straight into a paper.

A sequence table lists collection dates, total sizes, durations, and the rosbag format for several sequences. All five sequences were collected on 2021-12-23. Sizes run from 850M to 2.37G, and durations from 112 seconds to 349 seconds. The table gives concrete numbers about how big and how long the recordings are, listed in order from Seq1 onward. Storing each sequence as a rosbag matches how ROS users would consume it.

The README ends with an acknowledgement section. The authors express appreciation for the support of Shanghai West Hongqiao Navigation Technology Co., Ltd. That closing note sits alongside the citation request and the sequence information that make up the rest of the documentation.

• Source 1 · Project repository
• Source 2 · README