search

menandro / vfs

Variational Fisheye Stereo
49 stars 18 forks source link

Function time #4

Closed antithing closed 4 years ago

antithing commented 4 years ago

Hi, thanks again for making this code available. I have it running on windows, Cuda 10.0, and am testing the test_t265_data function with the provided data.

It runs, and looks good, but the function takes 15 seconds.

When I adjust the :

int nWarpIters = 20;    // Change to reduce processing time
int nSolverIters = 20;  // Change to reduce processing time

It takes 843 milliseconds.

Is this the expected time? Is it possible for this code to run in realtime?

thanks!

menandro commented 4 years ago

Yes. You’ll have to mix and match the nWarpIters, and nSolverIters. 848x800 is too big for realtime, but I have tested for 10fps and not so accurate.

The settings I use for my app is this:

float stereoScaling = 2.0f; float lambda = 5.0f; float nLevel = 5; float fScale = 2.0f; int nWarpIters = 5; int nSolverIters = 20; stereotgv->limitRange = 0.5f;

Lower lambda is smoother, but less accurate and needs higher nSolverIters. Lower lambda can be used for low nSolverIters. Low lambda and high nSolverIters is noisy.

Sent from my iPhone

On Mar 27, 2020, at 12:39 AM, antithing notifications@github.com wrote:

 Hi, thanks again for making this code available. I have it running on windows, Cuda 10.0, and am testing the test_t265_data function with the provided data.

It runs, and looks good, but the function takes 15 seconds.

When I adjust the :

int nWarpIters = 20; // Change to reduce processing time int nSolverIters = 20; // Change to reduce processing time It takes 843 milliseconds.

Is this the expected time? Is it possible for this code to run in realtime?

thanks!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

antithing commented 4 years ago

Thanks! that is faster.

My result (from the provided data) look like this:

depth

Quite different from the results on this page! Am I doing something wrong?

menandro commented 4 years ago

This is optical flow, so it looks correct.

Sent from my iPhone

On Mar 27, 2020, at 7:48 PM, antithing notifications@github.com wrote:

 Thanks! that is faster.

My result (from the provided data) look like this:

Quite different from the results on this page! Am I doing something wrong?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

antithing commented 4 years ago

Okay thanks! One last question... How can I get a greyscale depth map?

menandro commented 4 years ago

In the test_t265_data function, the 32-bit float depth map is copied to cv::Mat depth.