paulpanwang / cas6d Goto Github PK

Many thanks to the author for providing the code, which inspired my work. But there are errors in the network/cascade-refiner code. I have improved some errors here.
def update_state(self, scale, tranlation, quan):
self.scale_rmin = min(self.scale_rmin, scale.min() )...

But there is still an error in the code itself, which is about the parameters of tranlation and quan.
if not is_inference:
qn, sx, sy, sz, _ = vol_coords.shape
grids = pose_apply_th(que_imgs_info['poses_in'],
vol_coords.reshape(qn, sx * sy * sz, 3))
outputs['grids'] = grids
self.update_state(scale, tranlation, quan)###Tranlation and quan are not provided here
Does it represent the rotation and offset in the output dictionary?
I don't understand the actual meaning of translation and quan here. I hope the author can help improve the code. Thank very much，this determines whether this code can run.

Hi @paulpanwang, could you please provide the weights?