About the dexter+object dataset

[MengHao666]

Hi, I replement the DetNet in this paper, and get nearly same result as reported in paper, except the dexter+object dataset.

I think it might be due to something wrong in parsing dataset.My doubt is that ,in all other datasets(STB,RHD,ED), the 3D finger tip labels are provided in color camera, while dexter+object dataset provides them in depth camea. It seems that a transormation (transalation or rotation) is needed to transfer the labels to that in color camera.

And in the dataset, extrinsics and intrinsics of color camera , as well as the intrinsics of depth camera are provided. How should I do? Below I wrote my ugly code for parsing the dexter+object dataset, It would be very kind of u if anyone could check it. Anyway, just my 5 cents.

The parsing code of dexter+object dataset is on the below. Hope someone could help!
dexter_object.zip
.