Hello.
Thank you for your code, sincerely.
I have a problem with running the code for another dataset, ICL_NUIM dataset.
First, ICL_NUIM dataset provides intrinsic parameters(including focal length of pixel unit), camera pose(w2c coordinates, left-handed) and depth maps(16bit, scale factor = 5000). Scale factor means that if I want to get a depth for meter unit, just divide depth values by 5000. Following poses_bounds.npy format, I put them on my own poses_bounds.npy file. When putting on camera pose, I converted ICL_NUIM camera pose(wc2) to c2w using inverse function. I think that the pose matching was well done.
In this process, I have a question, how to get a right depth values? Using a depth map image(16bit, 0~255x255), I just took min and max values of the image. Then convert them to cm unit and m unit. The results are following.
Thank you for reading my question, and also for your nice code implementation, again.
Hello. Thank you for your code, sincerely. I have a problem with running the code for another dataset, ICL_NUIM dataset.
First, ICL_NUIM dataset provides intrinsic parameters(including focal length of pixel unit), camera pose(w2c coordinates, left-handed) and depth maps(16bit, scale factor = 5000). Scale factor means that if I want to get a depth for meter unit, just divide depth values by 5000. Following poses_bounds.npy format, I put them on my own poses_bounds.npy file. When putting on camera pose, I converted ICL_NUIM camera pose(wc2) to c2w using inverse function. I think that the pose matching was well done.
In this process, I have a question, how to get a right depth values? Using a depth map image(16bit, 0~255x255), I just took min and max values of the image. Then convert them to cm unit and m unit. The results are following.
Thank you for reading my question, and also for your nice code implementation, again.
This is for cm unit. cm
https://user-images.githubusercontent.com/98508690/184579316-a04ffd03-f292-4bf2-a326-1fb69970f3bd.mp4
This is for m unit. m
https://user-images.githubusercontent.com/98508690/184579250-dd91cb0e-ad5e-4887-bfcf-63c532036a40.mp4