gtCentroids =========== Each file contains individual ground truth points located near the center of the broccoli heads visible on each frame. The files are in plain text and use the following format: frame_20151104T123857.818990.pcd 0.42082009 0.13618046 0.69373178 ----------------1--------------- ----2----- ----3----- ----4----- 1) Is the file name of the point cloud frame. Each file is a frame grabbed by the sensor and the collection timestamp is used as its file name. Each file is in PCD format, which is the default format for the Point Cloud Library or PCL (https://pointclouds.org/). 2,3,4) The remaining three columns are the xyz coordinates of the ground truth point. The closest point to any xyz coordinate can be easily found to know where the broccoli heads are. gtClusters ========== Each file contains clusters of ground truth points that belong to the broccoli heads visible on each frame. The files are in plain text and use the following format: frame_20160426T144927.015924.pcd 0 23271 -0.052717373 -0.35826296 0.78500003 ----------------1--------------- -2- --3-- -----4------ -----5----- ----6----- 1) Is the file name of the point cloud frame. Each file is a frame grabbed by the sensor and the collection timestamp is used as its file name. Each file is in PCD format, which is the default format for the Point Cloud Library or PCL (https://pointclouds.org/). 2) Is the cluster number. All points in the same cluster are part of the same broccoli head. 3) Is the index number. This field is particularly useful for PCL users as many of PCL's algorithms handle and return indices. Indices are just a subset of the points of a cloud, not the points themselves, but just their index in the cloud. By using a point's index, its data can be retrieve in constant time. 4,5,6) The remaining three columns are the xyz coordinates of the ground truth point. Relevant publications ===================== * Hector A. Montes, Justin Le Louedec, Grzegorz Cielniak and Tom Duckett. Real-time detection of broccoli crops in 3D point clouds for autonomous robotic harvesting. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), October 25-29, 2020, Las Vegas, NV, USA . * Justin Le Louedec; Hector A. Montes; Tom Duckett; Grzegorz Cielniak. Segmentation and detection from organised 3D point clouds: a case study in broccoli head detection. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 14-19 June 2020, pp. 64-65, Seattle, WA, USA, USA. * K. Kusumam, T. Krajnik, S. Pearson, G. Cielniak and T. Duckett, 3D-vision based detection, localization, and sizing of broccoli heads in the field. Journal of Field Robotics, Wiley Online Library, 2017, 34, 1505-1518