v0.1.0

What's Changed

• Implement multiscale restarts by @eigenvivek in #1
• Set up ruff by @eigenvivek in #3
• Implement patient-agnostic training scripts by @eigenvivek in #5
• Add patient-specific training scripts by @eigenvivek in #6
• Move imports within functions by @eigenvivek in #7
• Add missing argument by @eigenvivek in #8
• Add Python API for registration by @eigenvivek in #9
• Refactor registration code by @eigenvivek in #10
• Replace setup.py with pyproject.toml by @eigenvivek in #11
• Implement modular initialization strategies by @eigenvivek in #12
• Fix bugs in fixed pose registration by @eigenvivek in #13
• Many miscellaneous updates by @eigenvivek in #14
• More miscellaneous updates by @eigenvivek in #15
• Compress images before saving by @eigenvivek in #16
• Bring animate API up to date by @eigenvivek in #17
• Fix directions for RAS+/LPS+ volumes by @eigenvivek in #18
• Add option to skip frames in xvr animate by @eigenvivek in #19
• Specifically reorient RAO images by @eigenvivek in #20
• Create LICENSE by @eigenvivek in #21
• Small utilities and bug fixes by @eigenvivek in #22
• Update LICENSE by @eigenvivek in #23
• Add all training and inference scripts by @eigenvivek in #24
• More user info by @eigenvivek in #25
• Update the demo by @eigenvivek in #26
• Switch to uv backend by @eigenvivek in #27
• Update README by @eigenvivek in #28
• Update README.md by @eigenvivek in #29
• Add a development guide by @eigenvivek in #30
• Add a roadmap by @eigenvivek in #32
• Bump DiffDRR version by @eigenvivek in #33
• Save poses in torch convention by @eigenvivek in #34
• Update colab tutorials by @eigenvivek in #35
• Make compute device specifiable by @eigenvivek in #36
• Add installation time by @eigenvivek in #37
• Add jupyter utility packages by @eigenvivek in #43
• Add an optional Dice loss to the pose regression network by @eigenvivek in #44
• Fix gradient accumulation in the trainer by @eigenvivek in #45
• Update restart command to use Trainer class by @eigenvivek in #46
• Add an optional weight to the Dice loss by @eigenvivek in #47
• Simplify chain training protocols by @eigenvivek in #48
• Start swapping scripts from python to bash by @eigenvivek in #49
• Add instructions to install pre-commit hooks locally by @eigenvivek in #50
• Update test-time optimization API by @eigenvivek in #51
• Add custom support for parameter categories by @eigenvivek in #52
• Sync with DiffDRR v0.6.0rc3 by @eigenvivek in #53
• Simplify package dependencies by @eigenvivek in #54
• Simplify environment and restart script by @eigenvivek in #55
• Add coordinate frame transforms during training by @eigenvivek in #56
• Refactor a separate io subpackage by @eigenvivek in #57
• Change default orientation to AP by @eigenvivek in #58
• Add LoRA finetuning option by @eigenvivek in #60
• Allow training to restart from a folder by @eigenvivek in #61
• Bump to latest version of DiffDRR by @eigenvivek in #63
• Update docstrings in README by @eigenvivek in #64
• Add option to disable the LR scheduler by @eigenvivek in #65
• Revert DiffDRR source to PyPI by @eigenvivek in #67
• Make options lowercase in help strings by @eigenvivek in #68
• Fix affine matrix inversion by @eigenvivek in #69
• Switch from pre-commit to prek by @eigenvivek in #71
• Speed up IO with torchio.SubjectsLoader by @eigenvivek in #70
• Add bash tags by @eigenvivek in #72
• Misc fixes in training loop by @eigenvivek in #73
• Get latest checkpoint by time modified by @eigenvivek in #74
• Re-weight the training loss function terms by @eigenvivek in #75
• Add a random crop sampler for large volumes by @eigenvivek in #76
• Bump xvr version by @eigenvivek in #77
• Simplify 2D augmentations by @eigenvivek in #78
• Refactor an unnecessary CPU-to-GPU transfer by @eigenvivek in #80
• Switch to glob patterns for training vols and masks by @eigenvivek in #81
• Add WandB run ID option for training restarts by @eigenvivek in #82
• Revert glob pattern for input volumes by @eigenvivek in #83
• Remove nonexistent module by @eigenvivek in #84
• Use same hyperparameters as v0.0.0rc1 by @eigenvivek in #85
• Pin DiffDRR version by @eigenvivek in #86
• Remove hf dependency with uvx by @eigenvivek in #87
• Remove LoRa finetuning capabilities by @eigenvivek in #88
• Get the last checkpoint by lexicographical name by @eigenvivek in #89
• Add pairwise multiview consistency loss by @eigenvivek in #90
• Hot-fix loss function by @eigenvivek in #91
• Update training hyperparameters by @eigenvivek in #92
• Initial docs commit by @eigenvivek in #96
• Refactor CLI by @eigenvivek in #97
• Fix uv environment setup by @eigenvivek in #99
• Lower non-zero image pixel threshold by @eigenvivek in #101
• Add histogram equalization by @eigenvivek in #102
• Add a restart-from-previous-run optimizer by @eigenvivek in #103
• Add antipodal pose estimation by @eigenvivek in #104
• Update scripts for new API by @eigenvivek in #105
• Update CLAHE probability by @eigenvivek in #106
• Add optional weighted sampling for multivolume training by @eigenvivek in #107
• Miscellaneous fixes by @eigenvivek in #108
• Save models more granularly during training by @eigenvivek in #109
• Add a single source of defaults for the API/CLI by @eigenvivek in #111
• Major update (v0.0.0rc2) by @eigenvivek in #113
• Change from zensical to mkdocs by @eigenvivek in #124
• Fix docs build by @eigenvivek in #125
• Update docs.yml by @eigenvivek in #126
• Test docs build during PRs by @eigenvivek in #128
• Add jaxtyping dependency by @eigenvivek in #157
• Add dataclass for camera intrinsics by @eigenvivek in #158
• Refactor registration module by @eigenvivek in #159
• Use attrs to simplify inheritance by @eigenvivek in #160
• Add some QoL to registration classes by @eigenvivek in #161
• Add fast rendering to training by @eigenvivek in #162
• Add init_only registration option by @eigenvivek in #163
• Switch to cyclopts backend by @eigenvivek in #164
• Fix singularities at zero by @eigenvivek in #165
• Deprecate metrics and visualization modules by @eigenvivek in #166
• Deprecate ANTs by @eigenvivek in #167
• Add Hausdorff distance to training losses by @eigenvivek in #169
• Modularize registration backend with protocols by @eigenvivek in #168
• Misc training fixes by @eigenvivek in #170
• A few miscellaneous registration improvements by @eigenvivek in #171
• Add methods to speed up finetuning by @eigenvivek in #172
• Compact experiments into a single folder by @eigenvivek in #173
• Add cnvx2 backend by @eigenvivek in #174
• Clean up data organization by @eigenvivek in #175
• Deprecate warp kwarg in training by @eigenvivek in #176
• Fix precision for cross-platform reproducibility by @eigenvivek in #177
• Change plateau threshold mode to abs by @eigenvivek in #178
• Reproduce main branch's registration results by @eigenvivek in #179
• Synchronize experiment scripts with main by @eigenvivek in #180
• Consolidate experiment files by @eigenvivek in #181
• Update experiments structure by @eigenvivek in #182
• Pin SO(3) parameterization by @eigenvivek in #183
• Add antipodal kwarg to model inference by @eigenvivek in #184
• Add runtime measurement and evaluation by @eigenvivek in #185
• Synchronize dev branch with main by @eigenvivek in #186
• Major upgrades to package structure by @eigenvivek in #187
• Migrate docs to zensical by @eigenvivek in #188
• Fix security vulnerabilities by @eigenvivek in #189
• Add radius flag to read_xray to isolate circular detector area by @Dark-Knight499 in #190
• Add test suite by @eigenvivek in #191
• Add citation by @eigenvivek in #193
• Add PyPI build and install instructions by @eigenvivek in #194
• Hotfix docs versioning by @eigenvivek in #195

New Contributors

• @Dark-Knight499 made their first contribution in #190

Full Changelog: https://github.com/eigenvivek/xvr/commits/v0.1.0