v0.3.10

What's Changed

• Move dims to device for torch.compile by @eigenvivek in #182
• Switch SE(3) backend to diffdrr.pose.RigidTransform by @eigenvivek in #185
• Add trilinear interpolation for rendering by @eigenvivek in #187

Full Changelog: v0.3.9...v0.3.10

v0.3.9

What's Changed

• Interactive pyvista plots by @eigenvivek in #174
• Add pytorchse3 by @eigenvivek in #176
• Support autodifferentiability for intrinsic parameters by @eigenvivek in #178
• Remove dependency on pytorch3d by @eigenvivek in #180

Full Changelog: v0.3.8...v0.3.9

Parsing intrinsic matrices and docs

What's Changed

• Add principal point from intrinsic matrix by @eigenvivek in #168
• Add 3D visualization example by @eigenvivek in #169
• Misc fixes by @eigenvivek in #170

Full Changelog: v0.3.7...v0.3.8

v0.3.7

What's Changed

• Add patch-based variants of image similarity metrics by @eigenvivek in #154
• Implement bone attenuation multiplier by @eigenvivek in #155
• Allow bone_attenuation_multiplier to be set once by @eigenvivek in #156
• Expose InstanceNorm eps argument by @eigenvivek in #160

Full Changelog: v0.3.6...v0.3.7

SE(3) and 3D plotting

What's Changed

• Support every SO(3) parameterization under the sun by @eigenvivek in #149
• Use conda for testing and docs deployment by @eigenvivek in #150
• Implement 3D visualization functions with pyvista by @eigenvivek in #151

Full Changelog: v0.3.5...v0.3.6

v0.3.5

Small tweaks to the detector (ensured rays only cast from source to target) and add a new image similarity metric.

What's Changed

• Cleanup experiment folder by @eigenvivek in #140
• Detector tweaks by @eigenvivek in #143
• Implement Gradient NCC by @eigenvivek in #146

Full Changelog: https://github.com/eigenvivek/DiffDRR/commits/v0.3.5

v0.3.3

Prioritize differentiable rendering

Final API change for DiffDRR

DRR objects now act purely as a differentiable renderer. That means DRR objects take in rotations and translations parameters to specify (batched) camera poses, and render X-rays from those poses.

The pose parameters are not stored as nn.Parameters of the module. This introduced too many annoyances with the DRR module, such as gradients not flowing backwards past the pose parameters, or only being able to render a specific batch size.

2D-to-3D registration is still possible with DiffDRR. This is done by creating a lightweight wrapper module that stores rotations and translations as nn.Parameters, and calls a DRR object in its forward call. An example is available in the updated documentation.

Small changes:

• DRR generation can now be performed patch-wise (i.e., generate a 600^2 DRR as four 300^2 in series). This means DiffDRR can generate insanely large DRRs since computation memory is no longer a constraint. See an example in the updated timing notebook.

What's Changed

• Differentiable rendering and large DRRs by @eigenvivek in #139

Full Changelog: v0.3.2...v0.3.3

v0.3.2

What's Changed

• Enable cuda testing with GitHub CI by @eigenvivek in #135
• New unified API for generating DRRs

The DRR module will now behave as follows:

• Initialize the DRR module as before (volume, spacing, height, pixel_spacing) except don't pass any pose parameters. Instead, initialize a set of dummy pose parameters that are internal to the model.
• Rename _update_params to move_carm. Change the function signature to input two batched vectors, one for translations and rotations. This will be the only way to reposition the C-Arm.
• To render a DRR, only use the forward call (i.e., drr()). No separate project function. This will return a batch of DRRs that are differentiable wrt to all downstream parameters.

Full Changelog: v0.3.1...v0.3.2

v0.3.1

What's Changed

• Add conversion to DeepDRR by @eigenvivek in #134

Full Changelog: v0.3.0...v0.3.1

v0.3.0

DiffDRR is now built using nbdev as the backend! Main advantage: there's now a functional docs site: https://vivekg.dev/DiffDRR

What's Changed

• Make default DiffDRR dtype torch.float32 by @eigenvivek in #123
• Implement with nbdev by @eigenvivek in #126

Full Changelog: v0.2.3...v0.3.0