failed to run tum demo

[ArthurHoumo]

Start GlORIE-SLAM at 2024-07-02 06:28:19, scene: tumrgbd-freiburg3_office, only_tracking: False, output: output/TUM_RGBD/demo/freiburg3_office

[INFO] Load droid pretrained checkpiont from ./pretrained/droid.pth! [INFO] Load ConvONet pretrained checkpiont from ./pretrained/middle_fine.pt! [MAPPER] Mapping Triggered! [TRACKER] Tracking Triggered! [TRACKER] : 100%|███████████████████████████████████████████████████████████████████| 100/100 [00:02<00:00, 33.54it/s] [TRACKER] Tracking Done! [MAPPER] Mapping Done! [TRACKER] Final Global BA Triggered! [TRACKER] Final Global BA Done! [POINTCLOUD] Updating pointcloud position ... [MAPPER] Number of pixels with valid droid depth: 174123/196608 (88.56%) Process Process-3: Traceback (most recent call last): File "/root/miniconda3/envs/glorie-slam/lib/python3.10/multiprocessing/process.py", line 314, in _bootstrap self.run() File "/root/miniconda3/envs/glorie-slam/lib/python3.10/multiprocessing/process.py", line 108, in run self._target(*self._args, *self._kwargs) File "/data/Arthur/open-projects/GlORIE-SLAM/src/slam.py", line 116, in mapping self.terminate() File "/data/Arthur/open-projects/GlORIE-SLAM/src/slam.py", line 135, in terminate self.mapper.final_refine(save_final_pcl=True) File "/data/Arthur/open-projects/GlORIE-SLAM/src/mapper.py", line 835, in final_refine self.mapping_keyframe(idx,video_idx,mono_depth,outer_joint_iters,num_joint_iters, File "/data/Arthur/open-projects/GlORIE-SLAM/src/mapper.py", line 719, in mapping_keyframe self.dynamic_r_add = self.dynamic_r_add/3.0 anchor_depth TypeError: unsupported operand type(s) for /: 'NoneType' and 'float'

GlORIE-SLAM finishes! +Style.RESET_ALL+2024-07-02 06:28:31

[zhangganlin]

Hi,

Thanks for pointing out the bug. It is because that in the tum-rgbd demo config, only the first 200 frames are used for simplicity (since it is just a demo config, only aim to give users a brief idea how our system works), which are not enough for the bootstrapping for the tracker, thus actually no frame is used for mapping (the mapper has not started yet), so in the refinement step in mapper, it cannot find any stored information about the dynamic_r.

Now I have updated the config to use the first 400 frames so that there are enough frames for the tracker for bootstrapping and it should work now.