Closed TianyuDai closed 2 years ago
Hi,
transform.txt
is the projection matrix, which comes from multiplying camera intrinsic (3x3 matrix) with the camera extrinsic ([rotation|translation], 3x4 matrix). .bin files
), .obj files
also store the uv coordinate and the vertex topology. Storing data in binary format provides greater space-efficiency and processing-efficiency. Thank you for the quick answer!
Thanks for sharing this dataset! I have read through DATASET_ASSET and DATA_STRUCTURE, but cannot find the information about camera matrices unit. I assume that the first three rows (3x3 matrix) in KRT is the camera intrinsic. Is this intrinsic in unit of pixel? What is the fourth row (0.0 0.0 0.0 0.0 0.0)? Are the last three rows (3x4 matrix) the camera extrinsic? Are the translation in extrinsic shown in unit of millimeter? Are vertices coordinates in xxx.obj and translation in xxx_transform.txt also saved in unit of millimeter? In addition, what is the dtype and exact meaning of the values in xxx.bin? I noticed that they are slightly more than the number of vertices saved in xxx.obj. Thank you!