It is trying to add two more dimensions for elementwise multiplication. I think there is no bug.

from mvsnet_pytorch.

It is trying to add two more dimensions for elementwise multiplication. I think there is no bug.

Thank you for you reply, However, the
depth_values = depth_values.view(*depth_values.shape, 1, 1)
what is the * mean?

from mvsnet_pytorch.

It is to unpack depth_values.shape as function arguments (you can refer to python tutorials for more information). It is the same as

depth_values = depth_values.view(depth_values.shape[0], depth_values.shape[1], 1, 1)  # only when len(depth_values.shape) == 2

from mvsnet_pytorch.

It is to unpack depth_values.shape as function arguments (you can refer to python tutorials for more information). It is the same as

depth_values = depth_values.view(depth_values.shape[0], depth_values.shape[1], 1, 1)  # only when len(depth_values.shape) == 2

Thank you ~ It works! And I want to konw whether the following code can replace this line?

p: probability volume [B, D, H, W]

depth_values: discrete depth values [B, D]

def depth_regression(p, depth_values):
#depth_values = torch.reshape(depth_values,(256))
depth_values = depth_values.cpu().numpy()
start = depth_values[0][0]
end = depth_values[0][-1]
interval = (end - start) / 256
disp = Variable(torch.Tensor(np.reshape(np.array(np.arange(start,end,interval)),[1,256,1,1])).cuda(), requires_grad=False)
disp = disp.repeat(p.size()[0],1,p.size()[2],p.size()[3])
#print('depth_value ', depth_values.shape)
depth = torch.sum(p * disp, 1)

return depth

from mvsnet_pytorch.

I'm not sure. You can check by yourself.

from mvsnet_pytorch.