Commit History

22 days ago

  1. README: which impact-reg-konfai versions read main

    Verified signature
    9307a78
  2. Each preset names its demo sample, a case of the fidopt corpus

    Verified signature
    4b7eec7

23 days ago

  1. Evaluation_with_seg: cast labels to int32 in the three presets of v1.8.7 too

    Verified signature
    6ffd54f
  2. Merge tag 'v1.8.7' into plan/main-presets⁠

    Verified signature
    268783a

24 days ago

  1. Evaluation_with_seg: cast labels to int32, not uint8⁠

    Verified signature
    31fcdb4

  1. The fidopt baselines move to their own repository, fideus-labs/fidopt

    Verified signature
    8c197ec
  2. No SynthMorph app: the FreeSurfer licence is not an open one

    Verified signature
    4634497
  3. The benchmark is fidopt

    Verified signature
    6f387be
  4. ConvexAdam_MIND: the authors' cost, balance and step⁠

    Verified signature
    1396c9d
  5. ConvexAdam_MIND as a preset of itk-impact's ConvexAdam at the authors' settings⁠

    Verified signature
    8b2856d
  6. The baselines run through impact-reg-konfai register, like the presets

    Verified signature
    f66d7c2
  7. The 12 registration baselines of ImagingBenchmarks as apps, next to the presets⁠

    Verified signature
    19c25c8

  1. build: pin itk-impact to 0.1.6 in the ConvexAdam presets⁠

    Verified signature
    09d2a19

  1. feat: presets for impact-reg-konfai 1.8.7, without integrator_n and with their model files⁠

    Verified signature
    efdeefe
  2. Restore 8acfab54 for impact-reg-konfai 1.8.6⁠

    Verified signature
    06efc39
  3. Presets for impact-reg-konfai 1.8.7: one IMPACT loss, sized from their declared cost

    Verified signature
    7fc77d1

  1. Write the configs in KonfAI 1.8.7's spellings⁠

    Verified signature
    8acfab5

  1. fix: drop ImpactWriteFeatureMaps, which elastix-IMPACT replaced⁠

    Verified signature
    2652081
  2. fix: the ConvexAdam presets run the settings ConvexAdam runs⁠

    Verified signature
    bd726b5
  3. feat: nine base presets, one per engine and per recommended pairing⁠

    Verified signature
    90b796d

  1. feat: no VRAM plan, every preset leaves its patch free and blends its tiles⁠

    Verified signature
    ae888a5

  1. Drop the config keys konfai 1.8 no longer reads⁠

    Verified signature
    9b707f7

  1. Write the registration result as Transform.h5 through the itktransform backend

    Verified signature
    b7287c9

  1. Drop the mask key from Patch, which konfai never binds⁠

    Verified signature
    fc7e61a
  2. A preset declares its displacement field, and only that⁠

    Verified signature
    ee3e448

  1. Add per-app icons and declare the registration task

    Verified signature
    3d2686d

  1. Require impact_reg_konfai 1.7.0 for the OME-Zarr displacement field

    Verified signature
    ab82a1c

  1. Pin itk-impact>=0.1.1 in the ConvexAdam presets for the Windows GIL fix

    Verified signature
    22cec16

  1. Drop the placeholder checkpoint; these engines are weightless⁠

    564cff2
  2. Move engine code to impact_reg_konfai.models; keep only config + weights⁠

    2f7c1ac

  1. fix(FireANTs): install fireants without its broken simpleitk pin; optional fused kernels (opt-in, pure-torch by default)

    Verified signature
    68e49c6

  1. Remove the CBCT_CT_TS preset

    1e7ef81

  1. Ignore __pycache__ and drop tracked bytecode

    a45bf82

  1. Add FireANTs presets (Affine/SyN/IMPACT); optional-mask defaults; elastix skips non-restrictive masks

    Verified signature
    6afc1ec

  1. fix(convexadam): disable only the TensorExpr fuser during registration⁠

    d2b63cb
  2. revert(convexadam): drop the legacy-JIT workaround, fixed model-side⁠

    86c3da0
  3. refactor(convexadam): scope the legacy-JIT-executor workaround to the registration⁠

    045bd3a
  4. fix(convexadam): disable JIT profiling executor for TorchScript feature models⁠

    cc45df7
  5. fix(convexadam): layers_mask as a 1/0 bitmask string, like elastix⁠

    9b3f426

  1. feat(convexadam): models as ModelSpec objects for ref/distance pickers⁠

    b22e3a0
  2. fix(convexadam): quiet the per-patch Registration bar in captured output⁠

    0a71243
  3. fix(convexadam): re-enable autograd for the itk-impact optimisation⁠

    173f532
  4. refactor: split elastix engine and type model parameters across presets⁠

    269b189

  1. Remove stale __pycache__/.pyc from preset bundles

    Verified signature
    6a3ecbf
  2. Update reg presets: mode knob + FOV evaluator + dimension-aware patch; models.json fetched from HF (unbundled)

    Verified signature
    442bdf8

  1. Add itk-impact ConvexAdam presets (coarse, fine, composite)

    Verified signature
    fd85e78

  1. Enable flip test-time augmentation on every preset

    7874a3b
  2. Show resolution level and metric in the registration progress bar

    2843777
  3. Add the generic uncertainty workflow to every preset

    8cce8eb
  4. Fix MRSeg IMPACT layer mask to match the model (1 layer)

    ae676a5