Hello Sir,
I have been trying to fine-tune D2Net on custom datasets, and from issue 60, it seems that we need to return exactly the same dictionary.
I have RGB, depth, pose, and intrinsic information of the robot. But I don't how to give bbox field. Could you tell me how to generate bbox from the above information?
Hello. We are using random 256x256 crops for training so the bbox are the coordinates of the top-left corner of the crop. If you are using full-resolution images then you can simply set this to 0, 0.
Hello Sir, I have been trying to fine-tune D2Net on custom datasets, and from issue 60, it seems that we need to return exactly the same dictionary. I have RGB, depth, pose, and intrinsic information of the robot. But I don't how to give
bbox
field. Could you tell me how to generatebbox
from the above information?