search

ECHO960 / PKU-MMD

Codes for PKU-MMD dataset.
Apache License 2.0
52 stars 8 forks source link

Camera calibration parameters #6

Open danasko opened 5 years ago

danasko commented 5 years ago

Hi, I'm wondering if you are able to provide your kinect v2 calibration parameters, since I'd like to produce point clouds from the dataset depth images. Thank you.

ECHO960 commented 5 years ago

hi, you can check this webpage for visualization: https://struct002.github.io/PKUMMD/

danasko commented 5 years ago

I'm sorry, but I did not find the calibration parameters, nor visualization code on the mentioned webpage. Could you specify what you meant by visualization?

ECHO960 commented 5 years ago

Hi, we include a camera matrix M in FAQ parts, but I am not sure where to get a matrix for each camera.

I am double-checking this with my team since I already graduated. Thanks

ECHO960 commented 5 years ago

Seems that we have 3 different sets of skeleton joints and depth/RGB images for each camera. Using M can map the skeleton joints to image coordinates, so M can be used for all three camera. Please note that we don't have real-world 3D coordinates for skeleton joints.

danasko commented 5 years ago

So I guess, since you don't have the intrinsic (focal length and principal point) and extrinsic (rotation matrix and translation vector) parameters of your camera, I won't be able to get point clouds in real-world coordinates from your depth data. Anyways, thank you for info.

ECHO960 commented 5 years ago

Well, the M I mentioned before is the camera intrinsic matrix. As for the extrinsic matrix, it is hard to obtain that as we slightly changed the camera position during capturing (for different sequences). So if you want to project the 3D points cloud, for the first view you can simply use M[I|0], but you need to estimate the R and t for the second and third view, sorry.

ECHO960 commented 5 years ago

wait, that intrinsic matrix is from RGB camera, let me see if I can get that from Depth camera

Hunger-Prevails commented 3 years ago

wait, that intrinsic matrix is from RGB camera, let me see if I can get that from Depth camera

Hi, the M you provided on https://struct002.github.io/PKUMMD/ was for the RGB cameras. Since I'm trying to project the camera-space 3D coordinates on to the respective depth images, I would rather need the intrinsic matrix for the depth cameras. I wonder how I can get them?

Thanks for maintaining this dataset btw, it's hugely helpful for my thesis