VisionWorks SFM Edit

Hi all,

New to the forum so thank you all in advance.

I am interesting in using the VisionWorks SFM samples provided by NVIDIA (thanks btw) as a platform for developing a SFM application.
The question is, has anyone developed the SFM to take individual frames rather than a video (i assume it breaks up the video into individual frames for processing anyway?). Each frame will be sequenced correctly so the software can use them in the correct order for point cloud generation.

If no one has, can anyone suggest where i can start to tackle this? I have experience with MatLab and C but I am only just starting out with C++ so any guidance welcome.



You can modify the main_sfm.cpp at ‘${HOME}/VisionWorks-SFM-0.90-Samples/samples/sfm/’ directly.

Currently, it is fixed to read the video file named as ‘parking_sfm.mp4’:

std::string sourceUri = app.findSampleFilePath("sfm/parking_sfm.mp4");

You can modify it to read the individual frames input for your use case.
// Check sample opencv_npp_interop located at ‘/usr/share/visionworks/sources/samples/opencv_npp_interop’ for the image input


Having some success so thank you AastaLLL much appreciated.