MIT-SPARK
segmentation fault when run ./testKimeraVIO
Description: I'didn't recive any error when I build the code, 100% at the end. but I got a segmentation fault when I test testKimeraVIO I don't know how to do, help me?
do anyone try a successful version or combination of it? please tell me.
Command:
cd build
./testKimeraVIO
Console output:
[ OK ] LCDFixture.processAndAddMonoFrame (208 ms)
[ RUN ] LCDFixture.processAndAddStereoFrame
I1209 21:14:52.626020 296818 LoopClosureDetector.cpp:70] LoopClosureDetector:: Loading vocabulary from ../tests/data/ForLoopClosureDetector/small_voc.yml.gz
I1209 21:14:52.630760 296818 LoopClosureDetector.cpp:73] Loaded vocabulary with 692 visual words.
I1209 21:14:52.856899 296818 LoopClosureDetector.cpp:194] LoopClosureDetector desctuctor called.
I1209 21:14:52.856983 296818 LcdThirdPartyWrapper.cpp:66] LcdThirdPartyWrapper desctuctor called.
[ OK ] LCDFixture.processAndAddStereoFrame (244 ms)
[ RUN ] LCDFixture.geometricVerificationCam2d2d
I1209 21:14:52.868907 296818 LoopClosureDetector.cpp:70] LoopClosureDetector:: Loading vocabulary from ../tests/data/ForLoopClosureDetector/small_voc.yml.gz
I1209 21:14:52.873562 296818 LoopClosureDetector.cpp:73] Loaded vocabulary with 692 visual words.
I1209 21:14:53.042711 296818 LoopClosureDetector.cpp:70] LoopClosureDetector:: Loading vocabulary from ../tests/data/ForLoopClosureDetector/small_voc.yml.gz
I1209 21:14:53.047205 296818 LoopClosureDetector.cpp:73] Loaded vocabulary with 692 visual words.
I1209 21:14:53.047345 296818 LoopClosureDetector.cpp:194] LoopClosureDetector desctuctor called.
I1209 21:14:53.047386 296818 LcdThirdPartyWrapper.cpp:66] LcdThirdPartyWrapper desctuctor called.
**Segmentation fault (core dumped)**
Additional files: Please attach all the files needed to reproduce the error.
Please give also the following information:
- Kimera-Vio branch, tag or commit used: master
- GTSAM version used: 4.2
- OpenGV version used: just clone from the source code but didn't change, that is master branch
- OpenCV version used: type
opencv_version: 3.4.0 - Operating system and version (e.g. Ubuntu 16.04 or Windows 10): Ubuntu 20.04
- Did you change the source code? (yes / no): yes I add #include <opencv2/viz.hpp> in include/frontend/Camera.h for solve some error during build I change all "Tracker" to "VIO::Tracker" in tests/testTracker.cpp for solving the definition of ambiguity, since there is a lib named Tracker in Opencv
and even I run another test example, ./scripts/stereoVIOEuroc.bash, it still occurs Segmentation fault, but there are no any clues for debugging. please help me
output in terminal like that
I1210 10:55:55.240368 21104 EurocDataProvider.cpp:89] Parsing Euroc dataset... W1210 10:55:55.695256 21104 EurocDataProvider.cpp:745] Value for final_k, 10000 is larger than total number of frames in dataset 2700 W1210 10:55:55.695338 21104 EurocDataProvider.cpp:748] Using final_k = 2700 I1210 10:55:55.874207 21104 RegularVioBackend.cpp:115] Using Regular VIO Backend. I1210 10:55:56.151775 21104 PipelineModule.h:451] MISO Pipeline Module: Display has no output queue registered. I1210 10:55:56.152212 21104 Pipeline.cpp:352] Pipeline Modules launched (parallel_run set to 1). I1210 10:55:56.153677 21186 Pipeline.cpp:102] Spinning Kimera-VIO. I1210 10:55:56.153764 21187 Pipeline.cpp:167] Shutting down VIO pipeline once processing has finished. I1210 10:55:56.153810 21187 Pipeline.cpp:180] I1210 10:55:56.163692 21185 EurocDataProvider.cpp:126] Running dataset between frame 50 and frame 2700 I1210 10:55:56.653965 21187 Pipeline.cpp:180] Statistics ----------- # log hz {avg +- std} [min, max] last data_provider_left_frame_queue Size [#] 3 5.7 {1.0 +- 0.0} [1.0, 1.0] 1.0 data_provider_right_frame_queue Size [#] 3 5.3 {1.0 +- 0.0} [1.0, 1.0] 1.0 frontend_input_queue Size [#] 2 2.8 {1.0 +- 0.0} [1.0, 1.0] 1.0 ./scripts/stereoVIOEuroc.bash: line 111: 21104 Segmentation fault (core dumped) $BUILD_PATH/stereoVIOEuroc --dataset_type="$DATASET_TYPE" --dataset_path="$DATASET_PATH" --initial_k=50 --final_k=10000 --params_folder_path="$PARAMS_PATH" --use_lcd="$USE_LCD" --vocabulary_path="$VOCABULARY_PATH/ORBvoc.yml" --flagfile="$PARAMS_PATH/flags/stereoVIOEuroc.flags" --flagfile="$PARAMS_PATH/flags/Mesher.flags" --flagfile="$PARAMS_PATH/flags/VioBackend.flags" --flagfile="$PARAMS_PATH/flags/RegularVioBackend.flags" --flagfile="$PARAMS_PATH/flags/Visualizer3D.flags" --logtostderr=1 --colorlogtostderr=1 --log_prefix=1 --v=0 --vmodule=Pipeline*=00 --log_output="$LOG_OUTPUT" --log_euroc_gt_data="$LOG_OUTPUT" --save_frontend_images=1 --visualize_frontend_images=1 --output_path="$OUTPUT_PATH"
today, I just run ./build/stereoVIOEuroc, plus some parameters in script, it could run, and when the --use_lcd=0 the Segmentation fault (core dumped) not happen again, like
./stereoVIOEuroc --dataset_type=0 --dataset_path=/mnt/d/zhipeng/datasets/Euroc_MAV/euroc-spark/MH_02_easy --initial_k=50 --final_k=10000 --params_folder_path=../params/Euroc --use_lcd=0 --vocabulary_path=$VOCABULARY_PATH/ORBvoc.yml --flagfile=../params/Euroc/flags/stereoVIOEuroc.flags --flagfile=../params/Euroc/flags/Mesher.flags --flagfile=../params/Euroc/flags/VioBackend.flags --flagfile=../params/Euroc/flags/RegularVioBackend.flags --flagfile=../params/Euroc/flags/Visualizer3D.flags --logtostderr=1 --colorlogtostderr=1 --log_prefix=1 --v=0 --vmodule=Pipeline*=00 --log_output=1 --log_euroc_gt_data=1 --save_frontend_images=1 --visualize_frontend_images=1 --output_path=../output_logs
if I set --use_lcd=1, that means loop closure detection is enable, then the Segmentation fault will happen again during the process running.
by debug, I find this happen in opengv, ransac module
then I check the git repository of https://github.com/laurentkneip/opengv/issues/88, and https://github.com/laurentkneip/opengv/issues/73
it seems opengv was built within CXX_STANDRD 11, but Kimera-VIO was built within 17. so I change the CMakeLists.txt in opengv, like set CXX_STANDARD 17 command the line 39 add_definitions (-march=native) that is for CXX_STANDARD 11
then, remove all the packages of opengv that have been installed, rebulid opengv and Kimera again. I could pass all the test when I run ./testKimeraVIO
more over the mapping procedure seems fail during the code running. is that the problem of configuration or just the algorithm?
