nnInteractive 1.1.3__tar.gz → 2.0.0__tar.gz

{nninteractive-1.1.3 → nninteractive-2.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nnInteractive
-Version: 1.1.3
+Version: 2.0.0
 Summary: Inference code for nnInteractive
 Author: Helmholtz Imaging Applied Computer Vision Lab
 Author-email: Fabian Isensee <f.isensee@dkfz-heidelberg.de>
@@ -219,8 +219,8 @@ Classifier: Topic :: Scientific/Engineering :: Medical Science Apps.
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: nnunetv2>=2.6
-Requires-Dist: torch<2.9.0,>=2.6
+Requires-Dist: nnunetv2>=2.7.0
+Requires-Dist: torch!=2.9.*,>=2.1.2
 Requires-Dist: acvl-utils<0.3,>=0.2.3
 Requires-Dist: batchgenerators>=0.25.1
 Provides-Extra: dev
@@ -236,19 +236,27 @@ Dynamic: license-file
 
 This repository provides the official Python backend for `nnInteractive`, a state-of-the-art framework for 3D promptable segmentation. It is designed for seamless integration into Python-based workflows—ideal for researchers, developers, and power users working directly with code.
 
-> ⚠️ **Temporary Compatibility Warning**  
-> There is a known issue with **PyTorch 2.9.0** causing **OOM errors during inference** in `nnInteractive` (related to 3D convolutions — see the PyTorch issue [here](https://github.com/pytorch/pytorch/issues/166122)).  
-> **Until this is resolved, please use PyTorch 2.8.0 or earlier.**
+`nnInteractive` is also available through graphical viewers (GUI) for those who prefer a visual workflow.
 
+### Recommended integrations (developed and maintained by our team)
 
-`nnInteractive` is also available through graphical viewers (GUI) for those who prefer a visual workflow. The napari and MITK integrations are developed and maintained by our team. Thanks to the community for contributing the 3D Slicer extension!
+<div align="center">
+
+| **<div align="center">[napari plugin](https://github.com/MIC-DKFZ/napari-nninteractive)</div>** | **<div align="center">[MITK integration](https://www.mitk.org/)</div>** |
+|-------------------|----------------------|
+| [<img src="imgs/Logos/napari.jpg" height="200">](https://github.com/MIC-DKFZ/napari-nninteractive) | [<img src="imgs/Logos/mitk.jpg" height="200">](https://www.mitk.org/) |
+
+</div>
+
+### Community-driven integrations
 
+Huge thanks to the community for contributing these integrations!
 
 <div align="center">
 
-| **<div align="center">[napari plugin](https://github.com/MIC-DKFZ/napari-nninteractive)</div>** | **<div align="center">[MITK integration](https://www.mitk.org/wiki/MITK-nnInteractive)</div>** | **<div align="center">[3D Slicer extension](https://github.com/coendevente/SlicerNNInteractive)</div>** | **<div align="center">[ITK-SNAP extension](https://itksnap-dls.readthedocs.io/en/latest/quick_start.html)</div>** | **<div align="center">[OHIF integration](https://github.com/CCI-Bonn/OHIF-AI)</div>** |
-|-------------------|----------------------|-------------------------|-------------------------|-------------------------|
-| [<img src="imgs/Logos/napari.jpg" height="200">](https://github.com/MIC-DKFZ/napari-nninteractive) | [<img src="imgs/Logos/mitk.jpg" height="200">](https://www.mitk.org/wiki/MITK-nnInteractive) | [<img src="imgs/Logos/3DSlicer.png" height="200">](https://github.com/coendevente/SlicerNNInteractive) | [<img src="imgs/Logos/snaplogo_sq.png" height="200">](https://itksnap-dls.readthedocs.io/en/latest/quick_start.html) | [<img src="imgs/Logos/ohif.png" height="200">](https://github.com/CCI-Bonn/OHIF-AI) |
+| **<div align="center">[3D Slicer extension](https://github.com/coendevente/SlicerNNInteractive)</div>** | **<div align="center">[ITK-SNAP extension](https://itksnap-dls.readthedocs.io/en/latest/quick_start.html)</div>** | **<div align="center">[OHIF integration](https://github.com/CCI-Bonn/OHIF-AI)</div>** |
+|-------------------------|-------------------------|-------------------------|
+| [<img src="imgs/Logos/3DSlicer.png" height="200">](https://github.com/coendevente/SlicerNNInteractive) | [<img src="imgs/Logos/snaplogo_sq.png" height="200">](https://itksnap-dls.readthedocs.io/en/latest/quick_start.html) | [<img src="imgs/Logos/ohif.png" height="200">](https://github.com/CCI-Bonn/OHIF-AI) |
 
 </div>
 
@@ -264,7 +272,6 @@ This repository provides the official Python backend for `nnInteractive`, a stat
 
 ---
 
-
 ## What is nnInteractive?
 
 > Isensee, F.\*, Rokuss, M.\*, Krämer, L.\*, Dinkelacker, S., Ravindran, A., Stritzke, F., Hamm, B., Wald, T., Langenberg, M., Ulrich, C., Deissler, J., Floca, R., & Maier-Hein, K. (2025). nnInteractive: Redefining 3D Promptable Segmentation. https://arxiv.org/abs/2503.08373 \
@@ -339,6 +346,9 @@ import SimpleITK as sitk
 from huggingface_hub import snapshot_download  # Install huggingface_hub if not already installed
 
 # --- Download Trained Model Weights (~400MB) ---
+# License reminder: The official nnInteractive checkpoint is licensed under
+# Creative Commons Attribution Non Commercial Share Alike 4.0 (CC BY-NC-SA 4.0).
+# See the License section of this readme!.
 REPO_ID = "nnInteractive/nnInteractive"
 MODEL_NAME = "nnInteractive_v1.0"  # Updated models may be available in the future
 DOWNLOAD_DIR = "/home/isensee/temp"  # Specify the download directory
@@ -360,7 +370,6 @@ session = nnInteractiveInferenceSession(
     verbose=False,
     torch_n_threads=os.cpu_count(),  # Use available CPU cores
     do_autozoom=True,  # Enables AutoZoom for better patching
-    use_pinned_memory=True,  # Optimizes GPU memory transfers
 )
 
 # Load the trained model

{nninteractive-1.1.3 → nninteractive-2.0.0}/nnInteractive/__init__.py

@@ -1,3 +1,3 @@
 from importlib.metadata import version as _version
 
-__version__ = _version("nnInteractive")
+__version__ = _version("nnInteractive")

{nninteractive-1.1.3 → nninteractive-2.0.0}/nnInteractive/inference/cvpr2025_challenge_baseline/predict.py

@@ -14,6 +14,7 @@ step (bbox prediction + 5 click refinements).  The evaluator will overwrite
 the same input file between calls, injecting updated clicks and the previous
 prediction (`prev_pred`).
 """
+
 from __future__ import annotations
 
 import argparse
@@ -28,11 +29,11 @@ from nnInteractive.inference.inference_session import nnInteractiveInferenceSess
 
 from nnunetv2.utilities.helpers import empty_cache
 
-
 # --------------------------------------------------------------------------- #
 #                           ===  EDIT BELOW  ===                              #
 # --------------------------------------------------------------------------- #
 
+
 def run_inference(
     image: np.ndarray,
     spacing: tuple[float, float, float],
@@ -66,27 +67,28 @@ def run_inference(
         classes start from 1 … N.  Make sure dtype is `np.uint8`.
     """
     session = nnInteractiveInferenceSession(
-        device=torch.device('cuda', 0),
+        device=torch.device("cuda", 0),
         use_torch_compile=False,
         verbose=False,
         torch_n_threads=os.cpu_count(),
         do_autozoom=True,
-        use_pinned_memory=True
     )
     session.initialize_from_trained_model_folder(
         model_training_output_dir=CHECKPOINT_DIR,
-        use_fold='all',
+        use_fold="all",
     )
     session.set_image(image[None].astype(np.float32))
-    target_buffer = torch.zeros(image.shape, dtype=torch.uint8, device='cpu')
+    target_buffer = torch.zeros(image.shape, dtype=torch.uint8, device="cpu")
     session.set_target_buffer(target_buffer)
     result = torch.zeros(image.shape, dtype=torch.uint8)
     num_objects = len(bbox) if bbox is not None else len(clicks)
     if bbox is not None and clicks is not None:
-        assert len(bbox) == len(clicks), ('Both bboxs and clicks lists are provided but with different length '
-                                          'suggesting different number of objects. This is not supported by this script '
-                                          'and it was not communicated by the organizing team that such cases exist '
-                                          'or how they are supposed to be handled.')
+        assert len(bbox) == len(clicks), (
+            "Both bboxs and clicks lists are provided but with different length "
+            "suggesting different number of objects. This is not supported by this script "
+            "and it was not communicated by the organizing team that such cases exist "
+            "or how they are supposed to be handled."
+        )
     for oid in range(1, num_objects + 1):
         # place previous segmentation
         if prev_pred is not None:
@@ -96,46 +98,50 @@ def run_inference(
         if bbox is not None:
             bbox_here = bbox[oid - 1]
             bbox_here = [
-                [bbox_here['z_min'], bbox_here['z_max'] + 1],
-                [bbox_here['z_mid_y_min'], bbox_here['z_mid_y_max'] + 1],
-                [bbox_here['z_mid_x_min'], bbox_here['z_mid_x_max'] + 1]
-                ]
+                [bbox_here["z_min"], bbox_here["z_max"] + 1],
+                [bbox_here["z_mid_y_min"], bbox_here["z_mid_y_max"] + 1],
+                [bbox_here["z_mid_x_min"], bbox_here["z_mid_x_max"] + 1],
+            ]
             session.add_bbox_interaction(bbox_here, include_interaction=True, run_prediction=False)
         if clicks is not None:
             clicks_here = clicks[oid - 1]
             clicks_order_here = clicks_order[oid - 1]
             fg_ptr = bg_ptr = 0
             for kind in clicks_order_here:
-                if kind == 'fg':
-                    click = clicks_here['fg'][fg_ptr]
+                if kind == "fg":
+                    click = clicks_here["fg"][fg_ptr]
                     fg_ptr += 1
                 else:
-                    click = clicks_here['bg'][bg_ptr]
+                    click = clicks_here["bg"][bg_ptr]
                     bg_ptr += 1
 
                 print(f"Class {oid}: {kind} click at {click}")
-                session.add_point_interaction(click, include_interaction=kind == 'fg', run_prediction=False)
+                session.add_point_interaction(click, include_interaction=kind == "fg", run_prediction=False)
         # now run inference on the last interaction center
         session.new_interaction_centers = [session.new_interaction_centers[-1]]
         session.new_interaction_zoom_out_factors = [session.new_interaction_zoom_out_factors[-1]]
         session._predict()
         result[session.target_buffer > 0] = oid
     del session
-    empty_cache(torch.device('cuda', 0))
+    empty_cache(torch.device("cuda", 0))
     return result.cpu().numpy()
 
+
 # --------------------------------------------------------------------------- #
 #                         ===  DO NOT EDIT BELOW ===                          #
 # --------------------------------------------------------------------------- #
 
+
 def parse_args() -> argparse.Namespace:
     p = argparse.ArgumentParser()
     p.add_argument("--case_path", required=True, help="Path to the input *.npz")
     p.add_argument("--save_path", required=True, help="Path to write output *.npz")
     return p.parse_args()
 
+
 # Adapt this to your checkpoint directory (relative to the script)
-CHECKPOINT_DIR = 'checkpoint_folder'
+CHECKPOINT_DIR = "checkpoint_folder"
+
 
 def main() -> None:
     args = parse_args()
@@ -147,12 +153,12 @@ def main() -> None:
 
     # ---------------------- Load input & prompts -------------------------- #
     data = np.load(case_path, allow_pickle=True)
-    image        = data["imgs"]
-    spacing      = tuple(data["spacing"])
-    bbox         = data.get("boxes")         # bounding boxes
-    clicks       = data.get("clicks")        # fg/bg clicks per class
+    image = data["imgs"]
+    spacing = tuple(data["spacing"])
+    bbox = data.get("boxes")  # bounding boxes
+    clicks = data.get("clicks")  # fg/bg clicks per class
     clicks_order = data.get("clicks_order")  # order of click types
-    prev_pred    = data.get("prev_pred")     # from last iteration
+    prev_pred = data.get("prev_pred")  # from last iteration
 
     # --------------------------- Inference -------------------------------- #
     seg = run_inference(image, spacing, bbox, clicks, clicks_order, prev_pred)
@@ -162,5 +168,6 @@ def main() -> None:
     np.savez_compressed(save_path, segs=seg.astype(np.uint8))
     print(f"[predict.py] Saved prediction to {save_path}")
 
+
 if __name__ == "__main__":
-    main()
+    main()