example run

.gitignore

@@ -0,0 +1 @@
+*nii.gz

README.md

@@ -15,7 +15,7 @@ datasets:
 
 # MindGrab (BrainChop MeshNet)
 
-MindGrab is a MeshNet-based skull-stripping model from the [BrainChop](https://github.com/neuroneural/brainchop) project. It takes 256×256×256 conformed T1 volumes and produces a binary brain mask. The checkpoint runs entirely in [tinygrad](https://github.com/tinygrad/tinygrad) and powers the in-browser BrainChop demos (WebGPU/WebGL).
+MindGrab is a MeshNet-based skull-stripping model from the [BrainChop](https://github.com/neuroneural/brainchop) project. It takes 256^3 conformed T1 volumes and produces a binary brain mask. The checkpoint runs entirely in [tinygrad](https://github.com/tinygrad/tinygrad) and powers the in-browser BrainChop demos (WebGPU/WebGL).
 
 ## Files
 
@@ -27,36 +27,25 @@ MindGrab is a MeshNet-based skull-stripping model from the [BrainChop](https://g
 
 Optionally, you can upload the WebGPU export (`net_mindgrab.js` + `net_mindgrab.safetensors`) from [`niivue-tinygrad`](https://github.com/spikedoanz/niivue-tinygrad) if you want to distribute browser-ready assets alongside the raw weights.
 
-## Usage
+Optionally, you can also run the model through the official frontend at [brainchop.org](https://brainchop.org/)
 
-```python
-from huggingface_hub import hf_hub_download
-from brainchop.tiny_meshnet import load_meshnet
 
-config_path = hf_hub_download("neuroneural/mindgrab", "model.json")
-weights_path = hf_hub_download("neuroneural/mindgrab", "model.pth")
+## Usage
 
-model = load_meshnet(config_path, weights_path)
-out = model(input_tensor)  # input_tensor: tinygrad Tensor (1,1,256,256,256)
-mask = out.argmax(axis=1)
+```sh
+uv pip install hf brainchop
 ```
 
-See `infer.py` in this repo for a full end-to-end example that loads a conformed T1 volume, runs MindGrab, and saves the mask as NIfTI.
-
-## Upload Instructions
+```python
+from pathlib import Path
 
-```bash
-huggingface-cli login
-huggingface-cli repo create mindgrab --type model --org neuroneural
-git clone git@hf.co:neuroneural/mindgrab
-cd mindgrab
+from huggingface_hub import hf_hub_download
+from brainchop import load, save, api
 
-# copy the files in this folder (model.json / layers.json / model.pth)
-cp /path/to/brainchop-cli/hf-model-mindgrab/* .
+model_dir = Path(hf_hub_download("neuroneural/mindgrab", "model.json")).parent
+hf_hub_download(repo_id, "model.pth")
 
-git add .
-git commit -m "Upload MindGrab MeshNet checkpoint"
-git push origin main
+vol = load("t1_crop.nii.gz")
+mask = api.segment(vol, str(model_dir))
+save(mask, "mindgrab_mask.nii.gz")
 ```
-
-Once published, link the model repo to the [MindGrab Space](https://huggingface.co/spaces/neuroneural/mindgrab) and the paper page on [hf.co/papers](https://huggingface.co/papers/submit) so people can discover the CLI, the Space, and the checkpoint from one central page.

infer.py

@@ -1,42 +0,0 @@
-"""Example inference using the MindGrab checkpoint on Hugging Face."""
-
-from pathlib import Path
-
-from huggingface_hub import hf_hub_download
-from tinygrad.tensor import Tensor
-
-from brainchop import Volume, load as bc_load, save
-from brainchop.tiny_meshnet import load_meshnet
-
-REPO_ID = "neuroneural/mindgrab"
-
-
-def download_model(repo_id: str = REPO_ID):
-    config_path = hf_hub_download(repo_id, "model.json")
-    weights_path = hf_hub_download(repo_id, "model.pth")
-    return config_path, weights_path
-
-
-def run_mindgrab(input_path: str, output_path: str = "mindgrab_mask.nii.gz") -> str:
-    config_path, weights_path = download_model()
-    model = load_meshnet(config_path, weights_path)
-
-    vol = bc_load(input_path)
-    input_tensor = vol.data.reshape(1, 1, 256, 256, 256).cast("float32")
-    logits = model(input_tensor)
-    mask = logits.argmax(axis=1).reshape(256, 256, 256).cast("uint8")
-
-    save(Volume(mask, vol.header), output_path)
-    return output_path
-
-
-if __name__ == "__main__":
-    import argparse
-
-    parser = argparse.ArgumentParser(description="Run MindGrab skull stripping")
-    parser.add_argument("input", help="Input conformed NIfTI volume (.nii/.nii.gz)")
-    parser.add_argument("-o", "--output", default="mindgrab_mask.nii.gz", help="Output mask path")
-    args = parser.parse_args()
-
-    output = run_mindgrab(args.input, args.output)
-    print(f"Saved mask to {output}")