1. The user own dataset should be in which folder catkin_ws/src folder or any other paths??

[jbehley]

1. The user own dataset should be in which folder catkin_ws/src folder or any other paths??

[jbehley]

To open a custom (user) dataset, you have to bring it into a compatible format: https://github.com/jbehley/point_labeler#folder-structure

See also our conversion script: https://github.com/jbehley/point_labeler/blob/master/scripts/kitti_raw2odometry.py for a dummy calib.txt.

[Psyclonus2887]

Is poses.txt necessary? I want to label the pointcloud files from KITTI, where I can not find poses.txt

[jbehley]

the poses.txt is necessary. You can either use a SLAM approach like SuMa (https://github.com/jbehley/SuMa) to estimate poses or convert poses from KITTI. See our conversion script kitti_raw2odometry.py.

[Psyclonus2887]

Thanks, I have run the conversion script and I'm sure the data set from KITTI is in compatible format. But when I run labeler.bin and tried to select the directory of my data set, the software just crushed and flashed back. I have no idea about where the problem is.

[jbehley]

Really hard to diagnose the problem like this. However, I would start by running the labeler with gdb and see where it crashes:

$ gdb -ex run ./labeler

and see what turns up when it crashes.

[Psyclonus2887]

Thank you. After using gdb, it seems the problem is about the bad lexical cast occurred when I select the directory of my dataset. But I still don't know what cause this.

[jbehley]

when you type backtrace or bt, you get the trace of function calls that came before the exception was thrown. Could you copy this?

However, I would guess that the calib.txtor poses.txt contain strings that can not be parsed at numbers. you should check that the numbers are formatted like 1.337 and not 1,337 or something like nan is inside the files.

[jbehley]

@Jessie20-20 could you solve your problems?

[Psyclonus2887]

Thank you! I can copy that and the problem sure is about calib.txt. According to the record, it was carrying out function KITTICalibration::initialize(const std::string& filename) in kitti_utils.cpp at line 50:
for (uint32_t i = 0; i < 12; ++i) { m(i / 4, i - int(i / 4) * 4) = boost::lexical_cast<float>(rv::trim(entries[i])); }
when the crash occurred. But I use the dummy calib.txt generated by kitti_raw2odometry.py, does it mean there are some errors in this python script? I don't find format problems in calib.txt as you said.

[Psyclonus2887]

OK I think I have found some of the reasons. The script kitti_raw2odometry.py uses code to generate calib.txt as follow:
calib_file = open(os.path.join(output_folder, "calib.txt"), "w")
calib_file.write("P0: 1 0 0 0 0 1 0 0 0 0 1 0\n")
calib_file.write("P1: 1 0 0 0 0 1 0 0 0 0 1 0\n")
calib_file.write("P2: 1 0 0 0 0 1 0 0 0 0 1 0\n")
calib_file.write("P3: 1 0 0 0 0 1 0 0 0 0 1 0\n")
calib_file.write("Tr: 1 0 0 0 0 1 0 0 0 0 1 0\n")
calib_file.close()
The punctuation : cause bad_lexical_cast in line 50.

[Psyclonus2887]

As for poses.txt, if I merely use the file generated by the script kitti_raw2odometry.py, some of the numbers can be written into pose.txt in the form of scientific notation with letter e, which can also cause bad_lexical_cast in line 98.
But after I transform all the numbers in correct form, the poses.txt still cause bad_lexical_cast... Really don't know what to do now.

[jbehley]

i think I cannot further any help on this as it seems to be a problem related to the parsing. ensure that your system locale is EN_US or something similar, since in some languages it is "3,0" instead of "3.0"; and this can cause some problems.

[Irving87]

hi, you can try
dos2unix filename
convert your calib.txt and poses.txt into unix format.