After using SFM pipeline, we want to transform the output into openMVS. But we came across a problem like this:
C:\openMVS\build\openMVG\Windows-AMD64-Release\Debug>openMVG_main_openMVG2openMVS -i C:/openMVS/build/openMVG/software/SfM/test_result/matches/sfm_data.json -d C:/Users/lzj/Desktop/MVS -o C:/Users/lzj/Desktop/MVS/scene.mvs
0% 10 20 30 40 50 60 70 80 90 100%
|----|----|----|----|----|----|----|----|----|----|
error: no image using camera 0 of platform 0
Scene saved to OpenMVS interface format:
22 images (0 calibrated)
0 Landmarks
OMG! We have spent two weeks on solving the problem, can you help us? we'll be very grateful!
The problem is that you are providing to the binary a sfm_data json that does not contains any camera poses.
the matches/sfm_data.json contains only the View and Intrinsic data.
Please use the sfm_data.bin that is stored in the SfM reconstruction directory (the one processed with Incremental or Global SfM)
Hi, pmoulon, thanks for your solution. But when i use sfm_data.bin, the error still remains. Only the Landmarks increases from 0 to 2885. I'm looking forward to your reply.
C:\openMVS\build\openMVG\Windows-AMD64-Release\Debug>openMVG_main_openMVG2openMVS -i C:/openMVS/build/openMVG/software/SfM/test_result/reconstruction_global/sfm_data.bin -d C:/Users/lzj/Desktop/MVS -o C:/Users/lzj/Desktop/MVS/scene.mvs
0% 10 20 30 40 50 60 70 80 90 100%
|----|----|----|----|----|----|----|----|----|----|
error: no image using camera 0 of platform 0
Scene saved to OpenMVS interface format:
22 images (0 calibrated)
2885 Landmarks
Can you share your sfm_data.bin file?
Of course, wait a minute please.
After investigation I think it's because the images cannot be reached.
What I suggest you to do is the following:
openMVG_main_ConvertSfM_DataFormat -i sfm_data.bin -o sfm_data.json
"root_path": "test_image",
to have the full_path where your image is and run again openMVG_main_openMVG2openMVS on your updated sfm_data.json
I will update the binary with an error message to display that the image corresponding to the view is not accessible.
The patch that will display you the appropriate message is now pushed to the develop
branch.
Hi, pmoulon, thanks for your solution, but i can't understand all of what you said. My steps are as follows:
(I use full_path
stands for C:/openMVS/build/openMVG/software/SfM
, because it's too long)
python SfM_GlobalPipeline.py full_path/test_image full_path/test_result
openMVG_main_ConvertSfM_DataFormat -i full_path/test_result/reconstruction_global/sfm_data.bin -o full_path/test_result/matches/sfm_data.json
openMVG_main_openMVG2openMVS -i full_path/test_result/matches/sfm_data.json -d C:/MVS -o C:/MVS/scene.mvs
I have two questions as follows:
edit the field "root_path": "test_image"
" as you said.I known maybe my questions is silly, Can you help me?.
As you say, lacking the full_path of test_image is the reason of my issue. But why the program run failed again after i use python SfM_GlobalPipeline.py full_path/test_image full_path/test_result
?
The pipeline should not fails with full_path.
Can you elaborate to tell me which part is not working?
On my computer I tried python SfM_GlobalPipeline.py ./ImageDataset_SceauxCastle/images/ ./reconstruction_foo
and everything ran fine.
python SfM_GlobalPipeline.py full_path/test_image full_path/test_result
run success.
but openMVG_main_openMVG2openMVS -i full_path/test_result/matches/sfm_data.json -d C:/MVS -o C:/MVS/scene.mvs
run faild. The progress bar is always 0%.
Please, note that you did not use a sfm_data json that contains camera poses!
The matches/sfm_data.json
contains only Views
and Intrinsics
information.
Please use the sfm_data.bin produced in full_path/test_result/reconstruction_global/sfm_data.bin
WOW, I have got the same problem,willing to watch how to solve it....
Please note that there is no problem:
1- Please use full path
2- Be sure that you are using a sfm_data that contains camera poses and structure (sfm_data.bin)
Now i run two program as follows:
Now The progress bar is in 30% after spending 10 minute, is it normal?
Dear pmoulon, thanks for your patience. Thank you very much!!
This problem has been solved! :)
Picture undistortion is not yet run in parallel but you can help on that ;-)
Please, can you close the issue?