This folder contains the dataset used in the IROS 2022 submitted paper "CosySlam: investigating object-level SLAM for detecting locomotion surfaces"

The trajectories are stored in rosbags. For each one, two rosbags are provided:   
- One with the images (/camera/color/image_raw), the IMU measurements (/camera/imu) and the groundtruth.  
- One with Cosypose's measurements (/cosypose) to simulate the behavior of a Cosypose rosnode, and IMU measurements (/camera/imu).  

Intel RealSense L515 and D435i RGB camera and internal IMU were used for the different datasets.

Trajectory descriptions and associated camera:  
- solo_approach: D435i embarked on Solo quadruped moving toward a stair step  
- solo_module: D435i embarked on Solo quadruped moving latteraly while looking at a module of steps  
- stairs_on_floor: D435i hand held moving around while looking at stairs laid on the floor  
- tless_circular: L515 hand held trajectory moving in circle around a table on which are laid T-LESS objects  
- tless_dynamic: Same as tless_circular with faster motions  


Groundtruth is provided thanks to our Qualysis mocap system which uses special message types "qualisys/Subject".   
The qualisys ROS package can be recovered from https://github.com/KumarRobotics/qualisys.   

Mocap topics are respectively called:  
- solo_approach and solo_module: /qualisys/Solo12_Top  
- stairs_on_floor: /qualisys/realsense_d435i  
- tless_circular and tless_dynamic: /qualysis/realsense  

The calibration files for both RealSense are provided in .json files. i_p_c and i_q_c stand for the extrinsic calibration between camera and IMU.