Open jediofgever opened 8 months ago
Hi. The bag-files are available at: http://subtdata.felk.cvut.cz/robingas/data/ Please, refer to the updated documentation for more details. Thanks.
Thank you for your reply. I read the paper, the trajectory optimization work which inherently uses traversablity estimation. I would like to perform binary classification of traversable / non-traversable comparison. What is the best performing model for this task ? Also is there hand labeled ground truths ? Or only “pseudo/proxy” labels. Lots of questions I know, thank you for your help!, I greatly appreciate it
Edit: in this link http://subtdata.felk.cvut.cz/robingas/data/traversability_estimation/TraversabilityDataset/supervised/clouds/
are the color points pseudo labels or networks predictions ?
Thanks for your interest. The dataset described at ./docs/trav_data.md contains manually annotated labels (binary traversability) for point clouds and RGB images. In addition, we release data sequences collected in a "self-supervised" way meaning that the labels are provided only for parts of a point cloud that was covered by a robot footprint path. Please, refer to the monoforce package for more details.
Thank you for the reply. I am not sure I understand yet. Could you please clarify what exactly is the color information on these point clouds http://subtdata.felk.cvut.cz/robingas/data/traversability_estimation/TraversabilityDataset/supervised/clouds/destaggered_points_colored/
Are they the ground truth labels or your method's estimation?
Hi there, I am also working on traversability estimation, and I would like to compare my method against this method.
I am running on ros2; I see that you have provided datasets as pcd and npy files. But I would like to run a bag file demo for traversability estimation. I would appreciate it if you could provide a few rosbags for my reference. Thank you