victorprad / InfiniTAM

A Framework for the Volumetric Integration of Depth Images
http://www.infinitam.org
Other
918 stars 351 forks source link

the result is abnormal when using saved rgb and depth images #85

Closed ThePetrelLHY closed 6 years ago

ThePetrelLHY commented 6 years ago

I can run InfiniTAMv2 with kinectv1 successfully. _20171024153933

The result is normal and i saved the rgb and depth images by tap key's' when running.

But when I used the saved images to run InfiniTAMv2, the result is abnormal and I noticed that the depth image is not like the depth image reading from kinectv1 directly. _20171024153911

Is there anything wrong or do i need to change some parameter to save the depth images?

AndreiBarsan commented 6 years ago

Hmm, the flat areas in the depth buffer visualization seem to point to a possible calibration issue. Are you able to run other pre-recorded sequences (like teddy) OK? What calibration parameters are you using at the moment?

ThePetrelLHY commented 6 years ago

Yeah, I can run the teddy ok... image

I just changed the rgb and depth intrinsics of the teddy calibration file using my kinect's parameters , and the remaining are unchanged like this image

I used the same calibration file when I runned the InfiniTAM using my kinect v1 directly and using the recorded sequences from the kinect... Maybe I also need to change the remaining parameters of the calibration file?

AndreiBarsan commented 6 years ago

Ah, you should try tuning the disparity calibration, i.e., the last two parameters. I'm not 100% sure that's what's causing the problem, but it looks like that to me.

Check out the 'readDisparityCalib' method (https://github.com/victorprad/InfiniTAM/blob/master/InfiniTAM/ITMLib/Objects/Camera/ITMCalibIO.cpp#L50) for more details on these parameters.

ThePetrelLHY commented 6 years ago

Oh! Thank you for your reply! That solves my problem! I try to delete the last two parameters and let disparity calibration use the defalut values, and it works!

Thank you very much!!!