-
Notifications
You must be signed in to change notification settings - Fork 5.9k
Closed
Labels
Description
System information (version)
- OpenCV => recent 3.4 ( cabad90 ) + contrib ( f9c8cb8 )
- Operating System / Platform => Ubuntu 18.04 (Aarch64) Jetson Nano
- Compiler => GCC 7.5.0
Detailed description
Test FastBilateralSolverTest.ReferenceAccuracy fails but hard to judge with human eys
$ ./opencv_test_ximgproc --gtest_filter=FastBilateralSolverTest.ReferenceAccuracy
:
[----------] 1 test from FastBilateralSolverTest
[ RUN ] FastBilateralSolverTest.ReferenceAccuracy
#iterations: 25
estimated error: 0.00147801
#iterations: 25
estimated error: 0.000920876
#iterations: 25
estimated error: 0.0010467
/opencv_contrib/modules/ximgproc/test/test_fbs_filter.cpp:107: Failure
Expected: (cvtest::norm(res, ref, NORM_INF)) <= (1), actual: 41 vs 1
[ FAILED ] FastBilateralSolverTest.ReferenceAccuracy (242 ms)
[----------] 1 test from FastBilateralSolverTest
:
[ FAILED ] 1 test, listed below:
[ FAILED ] FastBilateralSolverTest.ReferenceAccuracy
The test code is expecting bit-exact, but on platforms other than x86_64, I think bit-exact doesn't happen.
Loosening threshold is not really realistic, since diff 41 is bit too large (but this rarely happens)
I added a new comparison method in ts module, and modified to use that test when NOT on x86_64
I'll send a PR later
Steps to reproduce
Issue submission checklist
- I report the issue, it's not a question
- I checked the problem with documentation, FAQ, open issues,
answers.opencv.org, Stack Overflow, etc and have not found solution - I updated to latest OpenCV version and the issue is still there
- There is reproducer code and related data files: videos, images, onnx, etc


