SinglePixelCamera#

class deepinv.physics.SinglePixelCamera(m, img_shape, fast=True, device='cpu', dtype=torch.float32, rng=None, **kwargs)[source]#

Bases: DecomposablePhysics

Single pixel imaging camera.

Linear imaging operator with binary entries.

If fast=False, the operator uses a 2D subsampled hadamard transform, which keeps the first \(m\) modes according to the sequency ordering. In this case, the images should have a size which is a power of 2.

If fast=False, the operator is a random iid binary matrix with equal probability of \(1/\sqrt{m}\) or \(-1/\sqrt{m}\).

Both options allow for an efficient singular value decomposition (see deepinv.physics.DecomposablePhysics) The operator is always applied independently across channels.

It is recommended to use fast=True for image sizes bigger than 32 x 32, since the forward computation with fast=False has an \(O(mn)\) complexity, whereas with fast=True it has an \(O(n \log n)\) complexity.

An existing operator can be loaded from a saved .pth file via self.load_state_dict(save_path), in a similar fashion to torch.nn.Module.

Parameters:

  • m (int) – number of single pixel measurements per acquisition (m).

  • img_shape (tuple) – shape (C, H, W) of images.

  • fast (bool) – The operator is iid binary if false, otherwise A is a 2D subsampled hadamard transform.

  • device (str) – Device to store the forward matrix.

  • rng (torch.Generator) – (optional) a pseudorandom random number generator for the parameter generation. If None, the default Generator of PyTorch will be used.

Examples:

SinglePixelCamera operators with 16 binary patterns for 32x32 image:

>>> from deepinv.physics import SinglePixelCamera
>>> seed = torch.manual_seed(0) # Random seed for reproducibility
>>> x = torch.randn((1, 1, 32, 32)) # Define random 32x32 image
>>> physics = SinglePixelCamera(m=16, img_shape=(1, 32, 32), fast=True)
>>> torch.sum(physics.mask).item() # Number of measurements
48.0
>>> torch.round(physics(x)[:, :, :3, :3]).abs() # Compute measurements
tensor([[[[1., 0., 0.],
          [0., 0., 0.],
          [0., 0., 0.]]]])
U(x)[source]#

Applies the \(U\) operator of the SVD decomposition.

Note

This method should be overwritten by the user to define its custom DecomposablePhysics operator.

Parameters:

x (torch.Tensor) – input tensor

U_adjoint(x)[source]#

Applies the \(U^{\top}\) operator of the SVD decomposition.

Note

This method should be overwritten by the user to define its custom DecomposablePhysics operator.

Parameters:

x (torch.Tensor) – input tensor

V(y)[source]#

Applies the \(V\) operator of the SVD decomposition.

Note

This method should be overwritten by the user to define its custom DecomposablePhysics operator.

Parameters:

x (torch.Tensor) – input tensor

V_adjoint(x)[source]#

Applies the \(V^{\top}\) operator of the SVD decomposition.

Note

This method should be overwritten by the user to define its custom DecomposablePhysics operator.

Parameters:

x (torch.Tensor) – input tensor

Examples using SinglePixelCamera:#

A tour of forward sensing operators

A tour of forward sensing operators

PnP with custom optimization algorithm (Condat-Vu Primal-Dual)

PnP with custom optimization algorithm (Condat-Vu Primal-Dual)