Thanks for your wonderful work.
I tried adapting Tensor4D to the D-NeRF dataset presented in the paper, such as "Standup" and "Jumpingjacks". This involves converting between the DTU coordinate system and the Blender coordinate system. I obtained the following results (with feature_type=4d and iter=64k). I'm not sure if the issue is with my coordinate system conversion. If possible, could you provide the standard code for pose conversion between the two coordinate systems?