def inv2depth(inv_depth):
"""
Invert an inverse depth map to produce a depth map
Parameters
----------
inv_depth : torch.Tensor or list of torch.Tensor [B,1,H,W]
Inverse depth map
Returns
-------
depth : torch.Tensor or list of torch.Tensor [B,1,H,W]
Depth map
"""
if is_seq(inv_depth):
return [inv2depth(item) for item in inv_depth]
else:
return 1. / inv_depth.clamp(min=1e-6)
def inv2depth(self, inv_depth):
"""
Invert an inverse depth map to produce a depth map
Parameters
----------
inv_depth : torch.Tensor or list of torch.Tensor [B,1,H,W]
Inverse depth map
Returns
-------
depth : torch.Tensor or list of torch.Tensor [B,1,H,W]
Depth map
"""
if is_seq(inv_depth):
return [self.inv2depth(item) for item in inv_depth]
else:
return 1. * self.stereo_scale_factor / inv_depth
def resize_depth_preserve(depth, shape):
"""
Resizes depth map preserving all valid depth pixels
Multiple downsampled points can be assigned to the same pixel.
Parameters
----------
depth : np.array [h,w]
Depth map
shape : tuple (H,W)
Output shape
Returns
-------
depth : np.array [H,W,1]
Resized depth map
"""
# Return if depth value is None
if depth is None:
return depth
# If a single number is provided, use resize ratio
if not is_seq(shape):
shape = tuple(int(s * shape) for s in depth.shape)
# Store dimensions and reshapes to single column
depth = np.squeeze(depth)
h, w = depth.shape
x = depth.reshape(-1)
# Create coordinate grid
uv = np.mgrid[:h, :w].transpose(1, 2, 0).reshape(-1, 2)
# Filters valid points
idx = x > 0
crd, val = uv[idx], x[idx]
# Downsamples coordinates
crd[:, 0] = (crd[:, 0] * (shape[0] / h)).astype(np.int32)
crd[:, 1] = (crd[:, 1] * (shape[1] / w)).astype(np.int32)
# Filters points inside image
idx = (crd[:, 0] < shape[0]) & (crd[:, 1] < shape[1])
crd, val = crd[idx], val[idx]
# Creates downsampled depth image and assigns points
depth = np.zeros(shape)
depth[crd[:, 0], crd[:, 1]] = val
# Return resized depth map
return np.expand_dims(depth, axis=2)
def depth2inv(depth):
"""
Invert a depth map to produce an inverse depth map
Parameters
----------
depth : torch.Tensor or list of torch.Tensor [B,1,H,W]
Depth map
Returns
-------
inv_depth : torch.Tensor or list of torch.Tensor [B,1,H,W]
Inverse depth map
"""
if is_seq(depth):
return [depth2inv(item) for item in depth]
else:
inv_depth = 1. / depth.clamp(min=1e-6)
inv_depth[depth <= 0.] = 0.
return inv_depth