1ff17c74c37cf8decd86f695b5d2306f28165c65 take the second to last image of the vis images.
The last image might not be complete, because of Sigkilling the vis camera
56533980aba1bf92cc6712ab69a9990cdce8be22 only return the most recent events of a given photo tour.
All events are to much
8c1def63c52f16b99379a2bae1d32fe4f7b60b41 adding some logging to see, whether it is usefull to restart at that point, if no images were actually taken
468d048c140cf4994ccbf3efe62b5fa82db86ed5 updated Dockerfile to include opencv-emgu-net8 image.
This image can be built with the BuildEmgu.Dockerfile.
The image must then be tagged as opencv-emgu-net8, so that the actual Dockerfile of the Gateway server is able to pick it up
1bdb4f1b268e40c85ce0ddcbd1461c035f410394 added an integration test to test the virtual image worker image creation method.
Not a real integration though. DataContext is mocked
dff405031fee3704c11834364a1812744569c9bf opencv tests are now cross platform executable.
instead of imshow, ShowImage extension method is used, which checks whether the System is Win32 and has a GUI. If not, then no imshow is issued
4877a9538224b086ca1f7e21e69aee4320160452 another guard to sanity check the results of the opencv operations. I really do not know why, but Emgu or OpenCV has massive timing issues.
When sleeping for a period those errors vanish.
When executing only one test at a time, those errors are not existent. Only when multiple tests are run in parallel, then the issue can be reproduced.
4877a9538224b086ca1f7e21e69aee4320160452 another guard to sanity check the results of the opencv operations. I really do not know why, but Emgu or OpenCV has massive timing issues.
When sleeping for a period those errors vanish.
When executing only one test at a time, those errors are not existent. Only when multiple tests are run in parallel, then the issue can be reproduced.
Commit messages for #146
4877a9538224b086ca1f7e21e69aee4320160452 another guard to sanity check the results of the opencv operations. I really do not know why, but Emgu or OpenCV has massive timing issues.
When sleeping for a period those errors vanish.
When executing only one test at a time, those errors are not existent. Only when multiple tests are run in parallel, then the issue can be reproduced.
2e42d43b1e91bc669e2603d09a5acdbedd060381 this seems to fix the issue, that under linux in a docker container the cropping of vis images does not work all the time.
In OpenCV new Mats are allocated with heap data and only with a SetTo command, those Mats are actually initialized. Furthermore a mask should only have 1 channel. So I use another Mat as a mask
4877a9538224b086ca1f7e21e69aee4320160452 another guard to sanity check the results of the opencv operations. I really do not know why, but Emgu or OpenCV has massive timing issues.
When sleeping for a period those errors vanish.
When executing only one test at a time, those errors are not existent. Only when multiple tests are run in parallel, then the issue can be reproduced.
Commit messages for #146
4877a9538224b086ca1f7e21e69aee4320160452 another guard to sanity check the results of the opencv operations. I really do not know why, but Emgu or OpenCV has massive timing issues.
When sleeping for a period those errors vanish.
When executing only one test at a time, those errors are not existent. Only when multiple tests are run in parallel, then the issue can be reproduced.
2e42d43b1e91bc669e2603d09a5acdbedd060381 this seems to fix the issue, that under linux in a docker container the cropping of vis images does not work all the time.
In OpenCV new Mats are allocated with heap data and only with a SetTo command, those Mats are actually initialized. Furthermore a mask should only have 1 channel. So I use another Mat as a mask
Commit messages for #146
4877a9538224b086ca1f7e21e69aee4320160452 another guard to sanity check the results of the opencv operations. I really do not know why, but Emgu or OpenCV has massive timing issues.
When sleeping for a period those errors vanish.
When executing only one test at a time, those errors are not existent. Only when multiple tests are run in parallel, then the issue can be reproduced.
2e42d43b1e91bc669e2603d09a5acdbedd060381 this seems to fix the issue, that under linux in a docker container the cropping of vis images does not work all the time.
In OpenCV new Mats are allocated with heap data and only with a SetTo command, those Mats are actually initialized. Furthermore a mask should only have 1 channel. So I use another Mat as a mask
32504937affae1dbdb8bcc41853230ac87b9b81f Setting the values explicitly in the data matrix seems to always work. This means retrying is no longer necessary