Relocalization Error

[xls0Jacker]

When I tried to run VO using P001, I encountered the following problem:

(base) admin1@admin1-desktop:~/temp_5_14/catkin_ws$ roslaunch air_slam vo_tartanair.launch
... logging to /home/admin1/.ros/log/f3ede42a-31fa-11f0-b9ff-b1bfa00eb7ca/roslaunch-admin1-desktop-2191982.log
Checking log directory for disk usage. This may take a while.
Press Ctrl-C to interrupt
Done checking log file disk usage. Usage is <1GB.

started roslaunch server http://admin1-desktop:37935/

SUMMARY

PARAMETERS

• /rosdistro: noetic
• /rosversion: 1.17.2
• /visual_odometry/camera_config_path: /home/admin1/temp...
• /visual_odometry/config_path: /home/admin1/temp...
• /visual_odometry/dataroot: /home/admin1/temp...
• /visual_odometry/model_dir: /home/admin1/temp...
• /visual_odometry/saving_dir: /home/admin1/temp...

NODES
/
rviz (rviz/rviz)
visual_odometry (air_slam/visual_odometry)

auto-starting new master
process[master]: started with pid [2192014]
ROS_MASTER_URI=http://localhost:11311

setting /run_id to f3ede42a-31fa-11f0-b9ff-b1bfa00eb7ca
process[rosout-1]: started with pid [2192047]
started core service [/rosout]
process[visual_odometry-2]: started with pid [2192055]
process[rviz-3]: started with pid [2192056]
config_file = /home/admin1/temp_5_14/catkin_ws/src/configs/visual_odometry/vo_tartanair.yaml
config done
[INFO] [1747361436.299904761]: rviz version 1.14.26
[INFO] [1747361436.299973689]: compiled against Qt version 5.12.8
[INFO] [1747361436.299994196]: compiled against OGRE version 1.9.0 (Ghadamon)
[INFO] [1747361436.306385016]: Forcing OpenGl version 0.
map_builder done
dataset done
i ====== 0
i ====== 1
i ====== 2
i ====== 3
i ====== 4
i ====== 5
i ====== 6
i ====== 7
i ====== 8
i ====== 9
i ====== 10
i ====== 11
i ====== 12
i ====== 13
i ====== 14
i ====== 15
i ====== 16
i ====== 17
i ====== 18
i ====== 19
i ====== 20
i ====== 21
i ====== 22
i ====== 23
i ====== 24
i ====== 25
i ====== 26
i ====== 27
i ====== 28
i ====== 29
i ====== 30
i ====== 31
i ====== 32
i ====== 33
i ====== 34
i ====== 35
i ====== 36
i ====== 37
i ====== 38
i ====== 39
i ====== 40
i ====== 41
i ====== 42
i ====== 43
i ====== 44
i ====== 45
i ====== 46
i ====== 47
i ====== 48
i ====== 49
i ====== 50
i ====== 51
i ====== 52
i ====== 53
i ====== 54
i ====== 55
i ====== 56
i ====== 57
i ====== 58
i ====== 59
i ====== 60
i ====== 61
......
i ====== 425
i ====== 426
i ====== 427
i ====== 428
i ====== 429
i ====== 430
i ====== 431
i ====== 432
i ====== 433
Average FPS = -nan
Waiting to stop...
[INFO] [1747361436.864824676]: Stereo is NOT SUPPORTED
[INFO] [1747361436.864896183]: OpenGL device: NVIDIA GeForce RTX 3090/PCIe/SSE2
[INFO] [1747361436.864923574]: OpenGl version: 4.6 (GLSL 4.6).
Map building has been stopped
Save file to /home/admin1/temp_5_14/catkin_ws/src/debug/trajectory_v0.txt
trajectory.size = 0
map_path = /home/admin1/temp_5_14/catkin_ws/src/debug/AirSLAM_mapv0.bin
check map
_mappoints = 0
_keyframes = 0
Map saveing.....
Map saveing done!
[visual_odometry-2] process has finished cleanly
log file: /home/admin1/.ros/log/f3ede42a-31fa-11f0-b9ff-b1bfa00eb7ca/visual_odometry-2*.log

This is my dataroot, which has been adapted according to the EuRoC dataset.

And this is my modified launch file:

[hyeon-ji-shin]

Hi, I’m experiencing the same issue.
I didn’t encounter this issue when running on the EuRoC dataset, but I faced a similar error when trying it with a custom dataset.
Have you managed to resolve it?

[hyeon-ji-shin]

Hi, I gained some insight from your log files.

I noticed that in a normal run, the processing log looks like this:

i ====== 1513
One Frame Processing Time: 24 ms.
i ====== 1514
insert keyframe, id = 1508
One Frame Processing Time: 4 ms.

However, your log does not print the "One Frame Processing Time: ~ms" line, which seems to be caused by a failure to load the dataset.
https://github.com/sair-lab/AirSLAM/blob/619493b782db0dd1a768f5c7f8f9e09d1f7142b9/demo/visual_odometry.cpp#L40
As you can see in the link, if an error occurs during the load, the exception handling prevents the message from being printed.

I just managed to solve this issue.
In my case with a custom dataset, the timestamps for the left and right camera images did not perfectly match, unlike in the EuRoC dataset. The problem arose because the origin code attempts to fetch the right image using the exact timestamp of the left image, which caused a failure and prevented visualization in RViz.

(You can debug and edit codes for this problem with below.)
https://github.com/sair-lab/AirSLAM/blob/619493b782db0dd1a768f5c7f8f9e09d1f7142b9/src/dataset.cc#L9
https://github.com/sair-lab/AirSLAM/blob/619493b782db0dd1a768f5c7f8f9e09d1f7142b9/src/dataset.cc#L85

I know it's been a while since you posted this, but I'm sharing my solution in case anyone else runs into the same problem!