SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2201.03644 · 2022

3D Segmentation with Fully Trainable Gabor Kernels and Pearson's Correlation Coefficient

arXiv · PDF · Open in the Atlas

Code that ran

18 of 20 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; 0 honoured the contract we drafted, 0 violated it and 18 have no contract verdict; 0 failures of our instrument; 2 no recorded run), measured 2026-11-21; 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/2201.03644.json · badge: /api/ran/2201.03644.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
ibm/multimodal-3d-image-segmentation community (classifier label)
classifier basis not recorded on the link
18 ran · 2 no recorded run
FunctionStatusWhere it lives
corrcoef Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/custom_losses.py
code served (permissive licence) · get_code("7e174dc6b463a5df")
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 Tensor, ndim 2, dtype family float, shape (2, 4) 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 torch

torch.manual_seed(0)

# The function computes Pearson correlation coefficients
# It expects y_pred and y_true tensors with ndim in (3, 4, 5)
# The function reduces over dimensions [2:], so we need at least 3 dimensions
# batch_size=2, num_labels=4, and spatial dimensions for reduction

batch_size = 2
num_labels = 4
spatial_dim1 = 3
spatial_dim2 = 5

# Create 4D tensors (batch_size, num_labels, spatial_dim1, spatial_dim2)
y_pred = torch.randn(batch_size, num_labels, spatial_dim1, spatial_dim2)
y_true = torch.randn(batch_size, num_labels, spatial_dim1, spatial_dim2)

ARGS = {
    "y_pred": y_pred,
    "y_true": y_true,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) 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
corrcoef Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/tensorflow/nets/custom_losses.py
code served (permissive licence) · get_code("8c0f4545f7eb5aa7")
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 EagerTensor, ndim 2, dtype family float, shape (2, 5) 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 tensorflow as tf
import numpy as np

# Set seed for reproducibility
tf.random.set_seed(0)
np.random.seed(0)

# Create test tensors with shape (batch_size, spatial_dim1, spatial_dim2, num_labels)
# This is a 4D tensor as expected by the function (ndim=4 is valid per assertion)
batch_size = 2
spatial_dim1 = 3
spatial_dim2 = 4
num_labels = 5

y_true = tf.constant(
    np.random.randn(batch_size, spatial_dim1, spatial_dim2, num_labels).astype(np.float32)
)

y_pred = tf.constant(
    np.random.randn(batch_size, spatial_dim1, spatial_dim2, num_labels).astype(np.float32)
)

ARGS = {
    "y_true": y_true,
    "y_pred": y_pred,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (2, 5) — shape (2, 5) 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
dht2 Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/dht.py
code served (permissive licence) · get_code("a35f5fa643476fe2")
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 Tensor, ndim 3, dtype family float, shape (2, 4, 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 torch

torch.manual_seed(0)

# Create a small input tensor with shape (batch=2, height=4, width=3)
# The function applies FFT over the last two dimensions (-2, -1)
x = torch.randn(2, 4, 3)

# is_inverse parameter: test with False (forward transform)
is_inverse = False

ARGS = {
    'x': x,
    'is_inverse': is_inverse,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 4, 3) — shape (2, 4, 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
dht3 Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/dht.py
code served (permissive licence) · get_code("9cdf8a087accb760")
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 Tensor, ndim 4, dtype family float, shape (2, 4, 5, 6) 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 torch

torch.manual_seed(0)

# The function dht3 takes:
# - x: Input tensor (will be transformed over last 3 dimensions)
# - is_inverse: Boolean flag for inverse transform

# Create a small tensor with shape suitable for 3D transform
# Using distinct sizes: batch=2, then 3 spatial dims of sizes 4, 5, 6
x = torch.randn(2, 4, 5, 6)

ARGS = {
    'x': x,
    'is_inverse': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 4, 5, 6) — shape (2, 4, 5, 6) 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
dhtn Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/dht.py
code served (permissive licence) · get_code("bb9a2c9ca1e9ce68")
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 Tensor, ndim 3, dtype family float, shape (2, 4, 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 torch

torch.manual_seed(0)

# Create a small input tensor
# Shape: (batch=2, dim1=4, dim2=3) to have distinct dimensions
x = torch.randn(2, 4, 3)

# Dimensions to transform - can be a tuple of dimension indices
dim = (1, 2)

# Boolean flag for inverse transform
is_inverse = False

ARGS = {
    'x': x,
    'dim': dim,
    'is_inverse': is_inverse
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 4, 3) — shape (2, 4, 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_binary Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/experiments/metrics.py
code served (permissive licence) · get_code("69cccec0515e3734")
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

# Create two binary arrays for testing dice_binary
# y_true_bin: ground truth binary mask
# y_pred_bin: predicted binary mask

# Example 1: Some overlap between true and predicted
y_true_bin = np.array([1, 1, 0, 0, 1, 0], dtype=np.int32).reshape(2, 3)
y_pred_bin = np.array([1, 0, 0, 1, 1, 0], dtype=np.int32).reshape(2, 3)

ARGS = {
    "y_true_bin": y_true_bin,
    "y_pred_bin": y_pred_bin,
}
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
dice_coef Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/custom_losses.py
code served (permissive licence) · get_code("c0f7dcb15e62da45")
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 Tensor, ndim 2, dtype family float, shape (2, 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 torch

torch.manual_seed(0)

# Create small tensors for testing dice_coef
# The function expects y_pred and y_true to have ndim in (3, 4, 5)
# It sums over dimensions [2:], so we need at least 3 dimensions
# Let's use 4D tensors: (batch_size, num_labels, height, width)

batch_size = 2
num_labels = 3
height = 4
width = 5

# y_pred: prediction scores, shape (batch_size, num_labels, height, width)
y_pred = torch.rand(batch_size, num_labels, height, width)

# y_true: one-hot ground truth labels, same shape as y_pred
y_true = torch.randint(0, 2, (batch_size, num_labels, height, width)).float()

ARGS = {
    "y_pred": y_pred,
    "y_true": y_true,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 3) — shape (2, 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_coef Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/tensorflow/nets/custom_losses.py
code served (permissive licence) · get_code("a28ccf455345f250")
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 EagerTensor, ndim 2, dtype family float, shape (2, 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 tensorflow as tf
import numpy as np

# Set seed for reproducibility
tf.random.set_seed(0)

# Create small tensors for testing
batch_size = 2
num_labels = 3
spatial_dim = 4

# For a 3D case (ndim=3): shape should be (batch_size, spatial, num_labels)
y_true = tf.constant(
    np.random.randint(0, 2, size=(batch_size, spatial_dim, num_labels)),
    dtype=tf.float32
)

y_pred = tf.constant(
    np.random.uniform(0, 1, size=(batch_size, spatial_dim, num_labels)),
    dtype=tf.float32
)

ARGS = {
    "y_true": y_true,
    "y_pred": y_pred,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (2, 3) — shape (2, 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
get_reverse Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/hartley_operator.py
code served (permissive licence) · get_code("438db1ab4cce15b9")
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 Tensor, ndim 3, dtype family float, shape (2, 4, 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 torch

torch.manual_seed(0)

# Create a small input tensor
x = torch.randn(2, 4, 3)

# Define dims as a list of dimensions to reverse and roll
dims = [1, 2]

ARGS = {
    "x": x,
    "dims": dims
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 4, 3) — shape (2, 4, 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
get_spatial_padcrop Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/nets_utils.py
code served (permissive licence) · get_code("54c46b96c5bc7e74")
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 tuple, ndim 2, dtype family int; 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 torch

torch.manual_seed(0)

# Create a small tensor with batch and spatial dimensions
# Shape: (batch=2, channels=3, height=4, width=5)
x = torch.randn(2, 3, 4, 5)

# Define target shape for spatial dimensions only (height, width)
target_shape = [6, 7]

ARGS = {
    "x": x,
    "target_shape": target_shape,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, dtype family int — a tuple 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
grouping2d Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/hartley_mha.py
code served (permissive licence) · get_code("5eb6566103ab65ba")
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 Tensor, ndim 5, dtype family float, shape (2, 3, 24, 4, 2) 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 torch

torch.manual_seed(0)

# Construct arguments for grouping2d function
# x: Input tensor of shape (batch_size, num_heads, num_channels, height, width)
# patch_size: Tuple of (patch_h, patch_w)

batch_size = 2
num_heads = 3
num_channels = 4
height = 8
width = 6
patch_h = 2
patch_w = 3

x = torch.randn(batch_size, num_heads, num_channels, height, width)
patch_size = (patch_h, patch_w)

ARGS = {
    "x": x,
    "patch_size": patch_size,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 5, dtype family float, shape (2, 3, 24, 4, 2) — shape (2, 3, 24, 4, 2) 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
grouping3d Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/hartley_mha.py
code served (permissive licence) · get_code("19d38b4c1b50a3dd")
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 Tensor, ndim 6, dtype family float, shape (2, 3, 120, 4, 2, 2) 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 torch

torch.manual_seed(0)

# The function expects:
# - x: tensor of shape (batch_size, num_heads, num_channels, depth, height, width)
# - patch_size: tuple/list of 3 integers (patch_d, patch_h, patch_w)

# Create a small input tensor
# Using distinct sizes: batch=2, num_heads=3, channels=4, depth=8, height=6, width=10
batch_size = 2
num_heads = 3
num_channels = 4
depth = 8
height = 6
width = 10

x = torch.randn(batch_size, num_heads, num_channels, depth, height, width)

# Patch size must divide the spatial dimensions evenly
# depth=8 divisible by 2, height=6 divisible by 3, width=10 divisible by 5
patch_size = (2, 3, 5)

ARGS = {
    "x": x,
    "patch_size": patch_size
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 6, dtype family float, shape (2, 3, 120, 4, 2, 2) — shape (2, 3, 120, 4, 2, 2) 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
hartley_conv Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/hartley_operator.py
code served (permissive licence) · get_code("21a07b6a8cfc8b0c")
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 Tensor, ndim 2, dtype family float, shape (4, 2) 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 torch
import numpy as np

torch.manual_seed(0)

# Construct small tensors for Hartley convolution
# The function uses einsum with weight and x, so we need compatible shapes

# Create weight matrix: shape (4, 3) - represents a kernel
weight = torch.randn(4, 3)

# Create weight_reverse: weight with frequency axes reversed
# For a 2D tensor, reverse means flip along the last dimensions
weight_reverse = torch.flip(weight, dims=[-1])

# Create data x: shape (2, 3) - batch of 2, feature dim 3
x = torch.randn(2, 3)

# Create x_reverse: x with frequency axes reversed
x_reverse = torch.flip(x, dims=[-1])

# Create an einsum equation that works with these shapes
# weight: (4, 3), x or x±x_reverse: (2, 3)
# We want output in frequency domain
# A valid equation could be 'ij,kj->ik' which contracts on the j dimension
equation = 'ij,kj->ik'

ARGS = {
    'equation': equation,
    'weight': weight,
    'weight_reverse': weight_reverse,
    'x': x,
    'x_reverse': x_reverse
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (4, 2) — shape (4, 2) 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
normalize_modalities Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/experiments/utils.py
code served (permissive licence) · get_code("98138151a337f841")
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 float, shape (3, 4, 5) 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
import torch

torch.manual_seed(0)

# Create multichannel input data in channel-first format
# Shape: (num_channels, height, width) = (3, 4, 5)
data = np.random.randn(3, 4, 5).astype(np.float32)

# Define arguments for normalize_modalities
ARGS = {
    "data": data,
    "mask_val": None,
    "clip_val": None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — 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 float, shape (3, 4, 5) — shape (3, 4, 5) 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
pcc_loss Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/tensorflow/nets/custom_losses.py
code served (permissive licence) · get_code("21672f9602e64a31")
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 EagerTensor, ndim 1, dtype family float, shape (2,) 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 tensorflow as tf
import numpy as np

# Set seed for reproducibility
tf.random.set_seed(0)
np.random.seed(0)

# Create small test tensors
# y_true and y_pred should have shape (batch_size, spatial_dims..., num_labels)
# For 3D case: (batch_size, depth, height, width, num_labels)
batch_size = 2
num_labels = 4
spatial_dim = 3

# 3D case: shape (batch_size, depth, height, width, num_labels)
y_true = tf.constant(
    np.random.randn(batch_size, spatial_dim, spatial_dim, spatial_dim, num_labels),
    dtype=tf.float32
)

y_pred = tf.constant(
    np.random.randn(batch_size, spatial_dim, spatial_dim, spatial_dim, num_labels),
    dtype=tf.float32
)

ARGS = {
    "y_true": y_true,
    "y_pred": y_pred,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 1, dtype family float, shape (2,) — shape (2,) 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
spatial_padcrop Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/nets_utils.py
code served (permissive licence) · get_code("478c218f97db90f1")
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 Tensor, ndim 4, dtype family float, shape (2, 4, 6, 8) 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 torch

torch.manual_seed(0)

# Test case 1: 3D tensor (batch, channel, width) with padding needed
x_3d = torch.randn(2, 4, 5)
target_shape_3d = [8]

# Test case 2: 4D tensor (batch, channel, height, width) with cropping needed
x_4d = torch.randn(2, 4, 10, 12)
target_shape_4d = [6, 8]

# Test case 3: 5D tensor (batch, channel, depth, height, width) with mixed padding/cropping
x_5d = torch.randn(2, 4, 7, 9, 11)
target_shape_5d = [8, 10, 12]

# Use the 4D case as primary test
ARGS = {
    "x": x_4d,
    "target_shape": target_shape_4d,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 4, 6, 8) — shape (2, 4, 6, 8) 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
to_categorical Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/experiments/utils.py
code served (permissive licence) · get_code("c448fb68d0f2e025")
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 Tensor, ndim 5, dtype family float, shape (2, 5, 4, 3, 2) 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 torch

torch.manual_seed(0)

# Create a small label tensor with shape (B, 1, D, H, W)
# B=2 (batch), 1 (single label), D=4, H=3, W=2
y = torch.randint(0, 5, (2, 1, 4, 3, 2), dtype=torch.int32)

# num_classes can be inferred from max value or explicitly set
num_classes = 5

ARGS = {
    'y': y,
    'num_classes': num_classes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 5, dtype family float, shape (2, 5, 4, 3, 2) — shape (2, 5, 4, 3, 2) 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
upsampling Ran
ran on a synthesized input
ibm/multimodal-3d-image-segmentation/nets/architectures.py
code served (permissive licence) · get_code("e3e49350747f516f")
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 5, 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create a dict of tensors with different spatial sizes
# Key 0 is the reference (largest) tensor
# The function will upsample all tensors to match its spatial size
tensors = {
    0: torch.randn(2, 4, 8, 8),   # batch=2, channels=4, height=8, width=8 (reference/largest)
    1: torch.randn(2, 4, 4, 4),   # batch=2, channels=4, height=4, width=4 (smaller)
    2: torch.randn(2, 4, 2, 2),   # batch=2, channels=4, height=2, width=2 (smallest)
}

ARGS = {
    "tensors": tensors
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 5, 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
normalize_data Not yet run
unverified (harvested, no recorded run)
ibm/multimodal-3d-image-segmentation/experiments/utils.py
code served (permissive licence) · get_code("75b9413017c2330b")
How it ran

no invocation recorded.

ungrouping2d Not yet run
unverified (harvested, no recorded run)
ibm/multimodal-3d-image-segmentation/nets/hartley_mha.py
code served (permissive licence) · get_code("2ba406a7b5e08d61")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 20 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 20 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.

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 2026-11-21.

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 convolutional layer and loss function are two fundamental components in deep learning. Because of the success of conventional deep learning kernels, the less versatile Gabor kernels become less popular despite the fact that they can provide abundant features at different frequencies, orientations, and scales with much fewer parameters. For existing loss functions for multi-class image segmentation, there is usually a tradeoff among accuracy, robustness to hyperparameters, and manual weight selections for combining different losses. Therefore, to gain the benefits of using Gabor kernels while keeping the advantage of automatic feature generation in deep learning, we propose a fully trainable Gabor-based convolutional layer where all Gabor parameters are trainable through backpropagation. Furthermore, we propose a loss function based on the Pearson's correlation coefficient, which is accurate, robust to learning rates, and does not require manual weight selections. Experiments on 43 3D brain magnetic resonance images with 19 anatomical structures show that, using the proposed loss function with a proper combination of conventional and Gabor-based kernels, we can train a network with only 1.6 million parameters to achieve an average Dice coefficient of 83%. This size is 44 times smaller than the original V-Net which has 71 million parameters. This paper demonstrates the potentials of using learnable parametric kernels in deep learning for 3D segmentation.

For agents

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

get_harvested_code_for_paper("2201.03644")
get_code_for_paper("2201.03644")
have("2201.03644")

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

curl https://syntology.ai/api/ran/2201.03644.json

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

[![Syntology run record](https://syntology.ai/api/ran/2201.03644.svg)](https://syntology.ai/paper/2201.03644)

Connect an agent — have() is free.