SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1809.10486 · 2018

nnU-Net: Self-adapting Framework for U-Net-Based Medical Image Segmentation

arXiv · PDF · Open in the Atlas

Code that ran

10 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 7 have no contract verdict; 2 failures of our instrument; 1 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/1809.10486.json · badge: /api/ran/1809.10486.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
justld/nnunet_paddle community (classifier label)
classifier basis not recorded on the link
8 ran · 1 no recorded run
gift-surg/MONAIfbs community (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
FunctionStatusWhere it lives
create_data_list_of_dictionaries Ran
ran on a synthesized input
gift-surg/MONAIfbs/monaifbs/src/inference/monai_dynunet_inference.py
code served (permissive licence) · get_code("e02a343b07f81c0a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import tempfile

# Create temporary files for testing
temp_dir = tempfile.mkdtemp()
temp_file1 = os.path.join(temp_dir, "image1.nii.gz")
temp_file2 = os.path.join(temp_dir, "image2.nii.gz")

# Create actual files so os.path.isfile returns True
open(temp_file1, 'w').close()
open(temp_file2, 'w').close()

ARGS = {
    'input_files': [temp_file1, temp_file2]
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
create_data_list_of_dictionaries Ran
ran on a synthesized input
gift-surg/MONAIfbs/monaifbs/src/train/monai_dynunet_training.py
code served (permissive licence) · get_code("e3ab618a6509c413")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import tempfile

# Create a temporary directory and temporary files for testing
temp_dir = tempfile.mkdtemp()

# Create fake image and segmentation files
image_file_1 = os.path.join(temp_dir, "image_001.nii.gz")
seg_file_1 = os.path.join(temp_dir, "seg_001.nii.gz")
image_file_2 = os.path.join(temp_dir, "image_002.nii.gz")
seg_file_2 = os.path.join(temp_dir, "seg_002.nii.gz")

# Create the actual files (empty files are sufficient)
open(image_file_1, 'w').close()
open(seg_file_1, 'w').close()
open(image_file_2, 'w').close()
open(seg_file_2, 'w').close()

# Create the input file with two columns separated by comma
input_file = os.path.join(temp_dir, "data_list.txt")
with open(input_file, 'w') as f:
    f.write(f"{image_file_1},{seg_file_1}\n")
    f.write(f"{image_file_2},{seg_file_2}\n")

ARGS = {
    'input_file': input_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_foreground_mean Ran
ran on a synthesized input
justld/nnunet_paddle/nnunet_tools/ensemble.py
code served (permissive licence) · get_code("141eeed5ea28952a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float64, ndim 0, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct a results dictionary that matches the structure expected by get_foreground_mean
results = {
    'results': {
        'mean': {
            '0': {'Dice': 0.5},  # This key should be skipped
            '1': {'Dice': 0.85},
            '2': {'Dice': 0.90},
            '3': {'Dice': 0.88},
            'mean': {'Dice': 0.87}  # This key should be skipped
        }
    }
}

ARGS = {
    'results': results
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
jaccard Ran
ran, honoured the contract we drafted
justld/nnunet_paddle/nnunet_tools/metrics.py
code served (permissive licence) · get_code("12b5ba241c136bb3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create test and reference numpy arrays with matching shapes
test = np.array([
    [[1, 0, 1, 0],
     [0, 1, 0, 1]],
    [[1, 1, 0, 0],
     [0, 0, 1, 1]]
], dtype=np.int32)

reference = np.array([
    [[1, 0, 0, 0],
     [0, 1, 0, 0]],
    [[1, 0, 0, 0],
     [0, 0, 1, 0]]
], dtype=np.int32)

# Create a ConfusionMatrix instance
class ConfusionMatrix:
    def __init__(self, test=None, reference=None):
        self.tp = None
        self.fp = None
        self.tn = None
        self.fn = None
        self.size = None
        self.reference_empty = None
        self.reference_full = None
        self.test_empty = None
        self.test_full = None
        self.set_reference(reference)
        self.set_test(test)

    def set_test(self, test):
        self.test = test
        self.reset()

    def set_reference(self, reference):
        self.reference = reference
        self.reset()

    def reset(self):
        self.tp = None
        self.fp = None
        self.tn = None
        self.fn = None
        self.size = None
        self.test_empty = None
        self.test_full = None
        self.reference_empty = None
        self.reference_full = None

    def compute(self):
        if self.test is None or self.reference is None:
            raise ValueError("'test' and 'reference' must both be set to compute confusion matrix.")
        self.tp = int(((self.test != 0) * (self.reference != 0)).sum())
        self.fp = int(((self.test != 0) * (self.reference == 0)).sum())
        self.tn = int(((self.test == 0) * (self.reference == 0)).sum())
        self.fn = int(((self.test == 0) * (self.reference != 0)).sum())
        self.size = int(np.prod(self.reference.shape, dtype=np.int64))
        self.test_empty = not np.any(self.test)
        self.test_full = np.all(self.test)
        self.reference_empty = not np.any(self.reference)
        self.reference_full = np.all(self.reference)

    def get_matrix(self):
        for entry in (self.tp, self.fp, self.tn, self.fn):
            if entry is None:
                self.compute()
                break
        return (self.tp, self.fp, self.tn, self.fn)

    def get_size(self):
        if self.size is None:
            self.compute()
        return self.size

    def get_existence(self):
        for case in (self.test_empty, self.test_full, self.reference_empty, self.reference_full):
            if case is None:
                self.compute()
                break
        return (self.test_empty, self.test_full, self.reference_empty, self.reference_full)

confusion_matrix = ConfusionMatrix(test, reference)

ARGS = {
    'test': None,
    'reference': None,
    'confusion_matrix': confusion_matrix,
    'nan_for_nonexisting': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
pad_nd_image Ran
ran on a synthesized input
justld/nnunet_paddle/nnunet_tools/nnunet_infer.py
code served (permissive licence) · get_code("682d8e78b26b4ce6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 4, dtype family float, shape (2, 3, 128, 128) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a simple test image (4D array: batch, channel, height, width)
image = np.random.rand(2, 3, 64, 64).astype(np.float32)

# Define arguments for pad_nd_image
ARGS = {
    'image': image,
    'new_shape': (128, 128),
    'mode': 'constant',
    'kwargs': {'constant_values': 0},
    'return_slicer': False,
    'shape_must_be_divisible_by': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 4, dtype family float, shape (2, 3, 128, 128) — shape (2, 3, 128, 128) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
precision Ran
ran; our contract draft was wrong, not the code
justld/nnunet_paddle/nnunet_tools/metrics.py
code served (permissive licence) · get_code("035dc40fdafe0275")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create test and reference numpy arrays
# They need to be the same shape for the assertion to pass
test_array = np.array([[1, 0, 1, 0],
                       [1, 1, 0, 0]])

reference_array = np.array([[1, 0, 1, 1],
                            [1, 0, 0, 0]])

# Create a ConfusionMatrix instance
confusion_matrix_obj = None  # Will be None initially, letting the function create it

ARGS = {
    'test': test_array,
    'reference': reference_array,
    'confusion_matrix': confusion_matrix_obj,
    'nan_for_nonexisting': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
run_evaluation Ran
ran; our contract draft was wrong, not the code
justld/nnunet_paddle/nnunet_tools/evaluator.py
code served (permissive licence) · get_code("e7033836f7ce1958")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

class MockEvaluator:
    def __init__(self):
        self.labels = None
        self.test_data = None
        self.ref_data = None
    
    def set_test(self, test):
        self.test_data = test
    
    def set_reference(self, ref):
        self.ref_data = ref
    
    def construct_labels(self):
        self.labels = torch.tensor([0, 1, 0, 1])
    
    def evaluate(self, **metric_kwargs):
        return {
            'accuracy': 0.85,
            'precision': 0.88,
            'recall': 0.82
        }

test = "test_dataset_001"
ref = "reference_dataset_001"
evaluator = MockEvaluator()
metric_kwargs = {'threshold': 0.5, 'average': 'macro'}

ARGS = {
    'args': (test, ref, evaluator, metric_kwargs)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
subdirs Ran
ran on a synthesized input
justld/nnunet_paddle/nnunet_tools/postprocessing.py
code served (permissive licence) · get_code("dc6160cdd82098b8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import tempfile
import shutil

# Create a temporary directory structure for testing
temp_dir = tempfile.mkdtemp()

# Create some subdirectories with various names
os.makedirs(os.path.join(temp_dir, "subdir1"), exist_ok=True)
os.makedirs(os.path.join(temp_dir, "subdir2"), exist_ok=True)
os.makedirs(os.path.join(temp_dir, "prefix_test"), exist_ok=True)
os.makedirs(os.path.join(temp_dir, "test_suffix"), exist_ok=True)

# Create some files (which should be ignored)
with open(os.path.join(temp_dir, "file.txt"), "w") as f:
    f.write("test")

ARGS = {
    "folder": temp_dir,
    "join": True,
    "prefix": None,
    "suffix": None,
    "sort": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
subfiles Ran
ran on a synthesized input
justld/nnunet_paddle/nnunet_tools/postprocessing.py
code served (permissive licence) · get_code("f3e0abd3ec6c81be")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import tempfile
import shutil

# Create a temporary directory with some test files
temp_dir = tempfile.mkdtemp()

# Create some test files
test_files = [
    "file1.txt",
    "file2.txt",
    "data_001.csv",
    "data_002.csv",
    "readme.md",
    "config.json"
]

for filename in test_files:
    filepath = os.path.join(temp_dir, filename)
    with open(filepath, 'w') as f:
        f.write("test content")

# Define ARGS with all parameters for the subfiles function
ARGS = {
    "folder": temp_dir,
    "join": True,
    "prefix": None,
    "suffix": None,
    "sort": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
to_one_hot Ran
ran on a synthesized input
justld/nnunet_paddle/nnunet_tools/nnunet_fold_val.py
code served (permissive licence) · get_code("14526317bd565411")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family int, shape (3, 3, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a small segmentation map with distinct labels
seg = np.array([[0, 1, 2],
                [1, 0, 2],
                [2, 1, 0]], dtype=np.int32)

# Define all possible segment labels
all_seg_labels = np.array([0, 1, 2], dtype=np.int32)

ARGS = {
    "seg": seg,
    "all_seg_labels": all_seg_labels
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family int, shape (3, 3, 3) — shape (3, 3, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
dice Not yet run
unverified (harvested, no recorded run)
justld/nnunet_paddle/nnunet_tools/metrics.py
code served (permissive licence) · get_code("a85d8587c247d950")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 4 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 4 later) and 7 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
dice 4 other papers
4 later
2212.04497 (2022-12)
UNETR++: Delving into Efficient and Accurate 3D Medical Image Segmentation
2410.09299 (2024-10)
Hierarchical uncertainty estimation for learning-based registration in neuroimaging
2506.23468 (2025-06)
NavM rph: A Self-Evolving World Model for Vision-and-Language Navigation in Continuous Environments
and 1 more not named here
jaccard 4 other papers
4 later
2212.04497 (2022-12)
UNETR++: Delving into Efficient and Accurate 3D Medical Image Segmentation
2410.09299 (2024-10)
Hierarchical uncertainty estimation for learning-based registration in neuroimaging
2506.23468 (2025-06)
NavM rph: A Self-Evolving World Model for Vision-and-Language Navigation in Continuous Environments
and 1 more not named here
precision 4 other papers
4 later
2212.04497 (2022-12)
UNETR++: Delving into Efficient and Accurate 3D Medical Image Segmentation
2410.09299 (2024-10)
Hierarchical uncertainty estimation for learning-based registration in neuroimaging
2506.23468 (2025-06)
NavM rph: A Self-Evolving World Model for Vision-and-Language Navigation in Continuous Environments
and 1 more not named here
run_evaluation 2 other papers
2 later
2109.03201 (2021-09)
nnFormer: Interleaved Transformer for Volumetric Segmentation
2212.04497 (2022-12)
UNETR++: Delving into Efficient and Accurate 3D Medical Image Segmentation

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The U-Net was presented in 2015. With its straight-forward and successful architecture it quickly evolved to a commonly used benchmark in medical image segmentation. The adaptation of the U-Net to novel problems, however, comprises several degrees of freedom regarding the exact architecture, preprocessing, training and inference. These choices are not independent of each other and substantially impact the overall performance. The present paper introduces the nnU-Net ('no-new-Net'), which refers to a robust and self-adapting framework on the basis of 2D and 3D vanilla U-Nets. We argue the strong case for taking away superfluous bells and whistles of many proposed network designs and instead focus on the remaining aspects that make out the performance and generalizability of a method. We evaluate the nnU-Net in the context of the Medical Segmentation Decathlon challenge, which measures segmentation performance in ten disciplines comprising distinct entities, image modalities, image geometries and dataset sizes, with no manual adjustments between datasets allowed. At the time of manuscript submission, nnU-Net achieves the highest mean dice scores across all classes and seven phase 1 tasks (except class 1 in BrainTumour) in the online leaderboard of the challenge.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("1809.10486")
get_code_for_paper("1809.10486")
have("1809.10486")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/1809.10486.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/1809.10486.svg)](https://syntology.ai/paper/1809.10486)

Connect an agent — have() is free.