Initial upload via tools/push_to_hf.py (architecture: ilex.models.synthstrip.SynthStrip)

README.md

@@ -0,0 +1,66 @@
+---
+library_name: ilex
+tags:
+- jax
+- equinox
+- ilex
+- neuroimaging
+- skull
+license: mit
+license_link: https://surfer.nmr.mgh.harvard.edu/docs/synthstrip/
+---
+
+# SynthStrip -- Main model
+
+## Description
+
+SynthStrip skull-stripping network ported to JAX / Equinox from the FreeSurfer mri_synthstrip reference implementation. The network is trained with synthesis-based domain randomisation and is contrast-agnostic across modalities (T1w, T2w, FLAIR, DWI, etc.). It outputs a signed distance transform (SDT) of the brain surface; downstream tooling thresholds the SDT to obtain a binary brain mask.
+
+## Intended use
+
+General-purpose adult skull stripping. The 90% case.
+
+## Usage
+
+```python
+from ilex.models.synthstrip import SynthStrip
+model = SynthStrip.from_pretrained('ilex-hub/synthstrip.1')
+```
+
+## Authors
+
+Hoopes A., Mora J. S., Dalca A. V., Fischl B., Hoffmann M.
+
+## Citation
+
+Hoopes A., Mora J. S., Dalca A. V., Fischl B., Hoffmann M. (2022). SynthStrip: Skull-Stripping for Any Brain Image. NeuroImage, 260:119474. doi:10.1016/j.neuroimage.2022.119474
+
+### References
+
+- Hoopes A., Mora J. S., Dalca A. V., Fischl B., Hoffmann M. (2022). SynthStrip: Skull-Stripping for Any Brain Image. NeuroImage, 260:119474. doi:10.1016/j.neuroimage.2022.119474
+- Kelley W. et al. (2024). Boosting Skull-Stripping Performance for Pediatric Brain Images. IEEE International Symposium on Biomedical Imaging (ISBI).
+- Hoffmann M. (2025). Domain-Randomized Deep Learning for Neuroimage Analysis. IEEE Signal Processing Magazine.
+
+## License
+
+HF Hub license tag: `mit`
+
+**Effective terms:** MIT or CC-BY-4.0, at the licensee's option, per the upstream FreeSurfer offer at the license_url. The HF Hub tag is `mit` for searchability; the dual nature of the offer survives in this field and in the README License section.
+
+Upstream license reference: https://surfer.nmr.mgh.harvard.edu/docs/synthstrip/
+
+### Copyright
+
+Network architecture and training code: copyright (c) 2022 The General Hospital Corporation, distributed as part of FreeSurfer (GPL-2.0). Pretrained weights: dual-licensed by upstream as MIT or CC-BY-4.0 (see https://surfer.nmr.mgh.harvard.edu/docs/synthstrip/). JAX / Equinox port: copyright (c) the ilex authors, released under the Apache-2.0 / GPL-3.0 dual license used by ilex itself.
+
+## Upstream source
+
+Original weights / reference implementation: https://surfer.nmr.mgh.harvard.edu/docs/synthstrip/
+
+## Provenance
+
+This artefact was produced by [ilex](https://github.com/hypercoil/ilex)'s
+save/load pipeline. The architecture is implemented in
+`ilex.models.synthstrip.SynthStrip` and the weights have been converted
+from their upstream format. See the upstream source above
+for the canonical reference.

config.json

@@ -0,0 +1,49 @@
+{
+  "_ilex": {
+    "architecture": "ilex.models.synthstrip.model.SynthStrip",
+    "constructor_kwargs": {
+      "feat_mult": 2,
+      "inshape": null,
+      "max_features": 64,
+      "max_pool": 2,
+      "nb_conv_per_level": 2,
+      "nb_features": 16,
+      "nb_levels": 7,
+      "return_mask": false
+    },
+    "format": "ilex",
+    "framework_version": {
+      "equinox": "0.13.8",
+      "ilex": "0.0.0.dev0",
+      "jax": "0.10.0",
+      "jaxlib": "0.10.0",
+      "numpy": "2.4.4",
+      "safetensors": "0.7.0"
+    },
+    "has_state": false,
+    "origin": "ilex-native"
+  },
+  "authors": "Hoopes A., Mora J. S., Dalca A. V., Fischl B., Hoffmann M.",
+  "copyright": "Network architecture and training code: copyright (c) 2022 The General Hospital Corporation, distributed as part of FreeSurfer (GPL-2.0). Pretrained weights: dual-licensed by upstream as MIT or CC-BY-4.0 (see https://surfer.nmr.mgh.harvard.edu/docs/synthstrip/). JAX / Equinox port: copyright (c) the ilex authors, released under the Apache-2.0 / GPL-3.0 dual license used by ilex itself.",
+  "data_type": "nibabel",
+  "description": "SynthStrip skull-stripping network ported to JAX / Equinox from the FreeSurfer mri_synthstrip reference implementation. The network is trained with synthesis-based domain randomisation and is contrast-agnostic across modalities (T1w, T2w, FLAIR, DWI, etc.). It outputs a signed distance transform (SDT) of the brain surface; downstream tooling thresholds the SDT to obtain a binary brain mask.",
+  "equinox_version": "0.13.8",
+  "ilex_version": "0.0.0.dev0",
+  "image_classes": "Single-channel 3D structural MRI volume, contrast-agnostic. Inputs at 1 mm isotropic voxel size and conformed to LIA orientation are recommended for parity with the published checkpoint behaviour.",
+  "intended_use": "Research only. Skull stripping of 3D structural MRI volumes that have been conformed to 1 mm isotropic voxels, intensity-normalised (subtract minimum, divide by 99th percentile, clip to [0, 1]), and padded to a multiple of 64 along each spatial axis -- see the FreeSurfer mri_synthstrip preprocessing pipeline.",
+  "jax_version": "0.10.0",
+  "network_data_format": {
+    "inputs": {},
+    "outputs": {}
+  },
+  "numpy_version": "2.4.4",
+  "pred_classes": "Single-channel signed distance transform of the brain surface (positive outside, negative inside). Threshold at 1 mm to obtain the binary brain mask used by mri_synthstrip's CLI.",
+  "references": [
+    "Hoopes A., Mora J. S., Dalca A. V., Fischl B., Hoffmann M. (2022). SynthStrip: Skull-Stripping for Any Brain Image. NeuroImage, 260:119474. doi:10.1016/j.neuroimage.2022.119474",
+    "Kelley W. et al. (2024). Boosting Skull-Stripping Performance for Pediatric Brain Images. IEEE International Symposium on Biomedical Imaging (ISBI).",
+    "Hoffmann M. (2025). Domain-Randomized Deep Learning for Neuroimage Analysis. IEEE Signal Processing Magazine."
+  ],
+  "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20240725.json",
+  "task": "Skull stripping (brain extraction) of structural MRI volumes",
+  "version": "0.1.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5a6ce7a36c727b83a4ab6a0e50181dfa945c86d0f29d6f3a352031ebd681fda
+size 10271412