def backward(ctx, grad_dist1, grad_dist2):
opt = ctx.opt
p1, p2, dist1, dist2, idx1, idx2 = ctx.saved_tensors
grad_p1 = torch.zeros_like(p1)
grad_p2 = torch.zeros_like(p2)
if "cuda" in opt.device:
chamfer.backward(p1, p2, grad_p1, grad_p2, grad_dist1, grad_dist2,
idx1, idx2)
else:
raise NotImplementedError("CPU version not implemented")
return None, grad_p1, grad_p2
def backward(ctx, graddist1, graddist2, gradidx1, gradidx2):
xyz1, xyz2, idx1, idx2 = ctx.saved_tensors
graddist1 = graddist1.contiguous()
graddist2 = graddist2.contiguous()
gradxyz1 = torch.zeros(xyz1.size())
gradxyz2 = torch.zeros(xyz2.size())
gradxyz1 = gradxyz1.cuda()
gradxyz2 = gradxyz2.cuda()
chamfer.backward(xyz1, xyz2, gradxyz1, gradxyz2, graddist1, graddist2, idx1, idx2)
return gradxyz1, gradxyz2
def backward(ctx, grad_dist1, grad_dist2):
xyz1, xyz2, idx1, idx2 = ctx.saved_tensors
grad_xyz1, grad_xyz2 = chamfer.backward(xyz1, xyz2, idx1, idx2,
grad_dist1, grad_dist2)
return grad_xyz1, grad_xyz2