SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.12208 · CVPR · 2022

Self-supervised Learning of Adversarial Example: Towards Good Generalizations for Deepfake Detection

Yong Zhang, Yibing Song, Jue Wang, Lingqiao Liu, Liang Chen

arXiv · PDF · Open in the Atlas

Code that ran

27 of 36 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 2 violated it and 2 have no contract verdict; 23 failures of our instrument; 9 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/2203.12208.json · badge: /api/ran/2203.12208.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
liangchen527/SLADD community
Papers with Code archive
27 ran · 9 no recorded run
FunctionStatusWhere it lives
Block Ran
ran on a synthesized input · constructed an object; no result was computed
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("6acb761868805246")
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 an instance of its own class, Block: the run constructed the object and did not call it, so no result was computed. 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)

# Define ARGS dict with all parameters for Block.__init__
ARGS = {
    'in_filters': 32,
    'out_filters': 64,
    'reps': 3,
    'strides': 1,
    'start_with_relu': True,
    'grow_first': 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 Block — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Block 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
Xception Ran
ran on a synthesized input · constructed an object; no result was computed
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("7218169bb34fc5b4")
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 an instance of its own class, Xception: the run constructed the object and did not call it, so no result was computed. 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)

ARGS = {
    "num_classes": 2,
    "num_region": 7,
    "num_type": 2,
    "num_mag": 1,
    "inc": 6,
}
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 Xception — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Xception 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
_scale_array Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("1eba56f353d0bf37")
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 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 numpy as np

# Create a small numpy array with some values to test the function
arr = np.array([[10.5, 200, 50], [300, -5, 100]], dtype=np.float32)

ARGS = {
    "arr": arr,
    "clip": 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 ndarray, 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
apply_mask Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("9a6c144d88add499")
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 (4, 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 cv2
import numpy as np

torch_available = False
try:
    import torch
    torch_available = True
except ImportError:
    pass

# Create a small test image (3 channels, 4x3 dimensions)
img = np.random.randint(0, 256, size=(4, 3, 3), dtype=np.uint8)

# Create a mask with values in [0-255] range (4x3 dimensions to match image height/width)
mask = np.random.randint(0, 256, size=(4, 3), dtype=np.uint8)

ARGS = {
    "img": img,
    "mask": mask
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 int, shape (4, 3, 3) — shape (4, 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
blur_mask Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("e347cbbb16944cb7")
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 2, dtype family int, shape (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 numpy as np
import cv2

# Create a valid mask argument
# The function expects a mask that cv2.GaussianBlur can process
# GaussianBlur works with 2D or 3D numpy arrays (grayscale or color)
mask = np.random.randint(0, 256, size=(4, 3), dtype=np.uint8)

ARGS = {
    'mask': mask
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family int, shape (4, 3) — shape (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
channel_hist_match Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("fed5e64e52963ec7")
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 2, dtype family float, shape (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 numpy as np

# Create small test arrays for source and template
source = np.random.randint(0, 256, size=(4, 3), dtype=np.uint8).astype(np.float64)
template = np.random.randint(0, 256, size=(4, 3), dtype=np.uint8).astype(np.float64)

# Create a mask (optional parameter)
mask = np.random.rand(4, 3)

ARGS = {
    "source": source,
    "template": template,
    "hist_match_threshold": 255,
    "mask": mask
}
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 2, dtype family float, shape (4, 3) — shape (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
colorTransfer_avg Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("ba01fce9d6a52a04")
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 (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 numpy as np

torch_seed = 0

# Create small numpy arrays for source and target images
# Shape: (height, width, channels) with values in [0, 1]
img_src = np.random.rand(4, 3, 2).astype(np.float32)
img_tgt = np.random.rand(4, 3, 2).astype(np.float32)

# Create a mask (optional parameter)
# Same spatial dimensions as images, single channel
mask = np.ones((4, 3, 1), dtype=np.float32)

ARGS = {
    'img_src': img_src,
    'img_tgt': img_tgt,
    'mask': mask,
}
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 float, shape (4, 3, 2) — shape (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
colorTransfer_fs Ran
ran, violated the contract we drafted
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("e37a7fdfad057776")
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 (4, 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 sample image arrays in RGB format
# Images should be 2D arrays with shape (height, width, channels)
# Using small dimensions for testing

np.random.seed(0)

# Source image: small 4x3 RGB image (height=4, width=3, channels=3)
src_ = np.random.randint(0, 256, size=(4, 3, 3), dtype=np.uint8)

# Destination image: small 4x3 RGB image
dst_ = np.random.randint(0, 256, size=(4, 3, 3), dtype=np.uint8)

# Mask: binary mask with same spatial dimensions as images (4x3)
# Non-zero values indicate masked regions
mask = np.zeros((4, 3), dtype=np.uint8)
mask[1:3, 1:2] = 1  # Mark some pixels as part of the mask

ARGS = {
    "src_": src_,
    "dst_": dst_,
    "mask": mask
}
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 (4, 3, 3) — shape (4, 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
color_hist_match Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("c083a54678b93ce8")
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 (4, 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 small test images with 3 channels (RGB)
# Using distinct dimensions: height=4, width=3 to avoid ambiguity
height, width, channels = 4, 3, 3

# Source image: small random RGB image
src_im = np.random.randint(0, 256, size=(height, width, channels), dtype=np.uint8)

# Target image: small random RGB image with different distribution
tar_im = np.random.randint(50, 200, size=(height, width, channels), dtype=np.uint8)

# Optional mask: same spatial dimensions, single channel
mask = np.ones((height, width), dtype=np.float32)

# Parameters
hist_match_threshold = 255

ARGS = {
    'src_im': src_im,
    'tar_im': tar_im,
    'hist_match_threshold': hist_match_threshold,
    'mask': mask
}
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 (4, 3, 3) — shape (4, 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
color_transfer_idt Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("15355cf6f8ec7f19")
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 (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 numpy as np
import scipy as sp

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

# Create two small test images with shape (height, width, channels)
# Using distinct dimensions: 4x3x2 for i0 and 5x3x2 for i1
i0 = np.random.rand(4, 3, 2).astype(np.float32)
i0 = np.clip(i0, 0, 1)

i1 = np.random.rand(5, 3, 2).astype(np.float32)
i1 = np.clip(i1, 0, 1)

# Define the arguments dictionary
ARGS = {
    "i0": i0,
    "i1": i1,
    "bins": 256,
    "n_rot": 20,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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 (4, 3, 2) — shape (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
color_transfer_mix Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("92a49f601bcf78ae")
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 (4, 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
import cv2

# Create small test images with float values
# Image shape: (height, width, channels)
height, width, channels = 4, 3, 3

# Create source image with random float values in range [0, 1]
img_src = np.random.rand(height, width, channels).astype(np.float32)

# Create target image with random float values in range [0, 1]
img_trg = np.random.rand(height, width, channels).astype(np.float32)

ARGS = {
    'img_src': img_src,
    'img_trg': img_trg,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 float, shape (4, 3, 3) — shape (4, 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
color_transfer_mkl Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("774c440758cf4d89")
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 (4, 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 two small color images (height x width x channels)
# Using distinct dimensions to make shapes unambiguous
np.random.seed(0)

# First image: 4x3 with 3 color channels
x0 = np.random.rand(4, 3, 3).astype(np.float32)
x0 = np.clip(x0, 0, 1)

# Second image: 5x4 with 3 color channels (different spatial dims)
x1 = np.random.rand(5, 4, 3).astype(np.float32)
x1 = np.clip(x1, 0, 1)

ARGS = {
    "x0": x0,
    "x1": x1,
}
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 float, shape (4, 3, 3) — shape (4, 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
color_transfer_sot Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("8f4b4f9c257b13ed")
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 (4, 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
import cv2

# Create small test images as float arrays
# src and trg must be float, same shape, rank 3 (h, w, c)
np.random.seed(0)

height = 4
width = 3
channels = 3

# Create source image as float32
src = np.random.rand(height, width, channels).astype(np.float32)

# Create target image as float32 with same shape
trg = np.random.rand(height, width, channels).astype(np.float32)

ARGS = {
    'src': src,
    'trg': trg,
    'steps': 2,
    'batch_size': 2,
    'reg_sigmaXY': 16.0,
    'reg_sigmaV': 5.0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 float, shape (4, 3, 3) — shape (4, 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
copy_fake_to_real Ran
ran, violated the contract we drafted
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("5319944dba530319")
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 (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 numpy as np

# Analyze the function:
# - realimg: numpy array, used in arithmetic operations
# - fakeimg: numpy array, used in arithmetic operations
# - mask: numpy array, expanded along axis 2
# - mag: scalar, used in arithmetic operations

# The function performs element-wise operations on arrays.
# mask is expanded to have shape (..., ..., 1) via expand_dims
# All arrays should be compatible for broadcasting

# Create small distinct-sized arrays
realimg = np.random.rand(2, 4, 3).astype(np.float32)  # shape: (2, 4, 3)
fakeimg = np.random.rand(2, 4, 3).astype(np.float32)  # shape: (2, 4, 3)
mask = np.random.rand(2, 4).astype(np.float32)        # shape: (2, 4), will be expanded to (2, 4, 1)
mag = 0.5                                              # scalar magnitude

ARGS = {
    "realimg": realimg,
    "fakeimg": fakeimg,
    "mask": mask,
    "mag": mag
}
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 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
correct_colours Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("b85534f53cf3248a")
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 (100, 100, 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
import cv2

# Create small test images (BGR format, typical for cv2)
# Image 1: 100x100 RGB image
im1 = np.random.randint(0, 256, (100, 100, 3), dtype=np.uint8)

# Image 2: 100x100 RGB image
im2 = np.random.randint(0, 256, (100, 100, 3), dtype=np.uint8)

# Create landmarks for face detection (68 points in dlib format)
# Points 36-41 are right eye, 42-47 are left eye
# We need at least 48 points to access indices up to 47
landmarks1 = np.random.rand(68, 2) * 100

# Ensure the eye points are reasonably spaced for blur calculation
# Right eye points (36-41)
landmarks1[36:42] = np.array([[20, 30], [25, 28], [30, 30], [32, 35], [28, 38], [22, 36]])
# Left eye points (42-47)
landmarks1[42:48] = np.array([[60, 30], [65, 28], [70, 30], [72, 35], [68, 38], [62, 36]])

ARGS = {
    "im1": im1,
    "im2": im2,
    "landmarks1": landmarks1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 (100, 100, 3) — shape (100, 100, 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
dist Ran
ran; our contract draft was wrong, not the code
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("669a43459c1dcc98")
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 math

def dist(p1, p2):
    return math.sqrt((p1[0] - p2[0]) ** 2 + (p1[1] - p2[1]) ** 2)

# p1 and p2 are expected to be tuple-like objects with at least 2 elements
# representing 2D coordinates (x, y)
p1 = (0, 0)
p2 = (3, 4)

ARGS = {
    "p1": p1,
    "p2": p2,
}
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 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
generate_random_mask Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("0ced88afb54982e1")
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 2, dtype family float, shape (256, 256) 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
from skimage import measure

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

# Create a sample mask - a 2D numpy array with some non-zero values
# This represents a binary mask or probability mask
mask = np.zeros((256, 256))
# Add a small region with values > 0.1 so regionprops finds at least one region
mask[100:150, 100:150] = 0.5

# Define arguments
ARGS = {
    "mask": mask,
    "res": 256
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, skimage — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (256, 256) — shape (256, 256) 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_five_key Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("95234979fc3cc146")
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 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 numpy as np

# Create a 68x2 array of landmarks (68 facial landmarks with x,y coordinates)
landmarks_68 = np.random.rand(68, 2).astype('float32') * 100

ARGS = {
    'landmarks_68': landmarks_68
}
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 list, ndim 2, dtype family int — 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
lab_image_stats Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("4de3aeeb2b8d9248")
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 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 cv2
import numpy as np

# Create a valid LAB image for the function
# LAB images are typically represented as uint8 arrays with shape (height, width, 3)
# Create a small 4x3 LAB image
lab_image = np.zeros((4, 3, 3), dtype=np.uint8)

# Fill with some sample LAB values
# L channel: 0-255, a channel: 0-255, b channel: 0-255
lab_image[:, :, 0] = np.random.randint(0, 256, (4, 3), dtype=np.uint8)  # L channel
lab_image[:, :, 1] = np.random.randint(0, 256, (4, 3), dtype=np.uint8)  # a channel
lab_image[:, :, 2] = np.random.randint(0, 256, (4, 3), dtype=np.uint8)  # b channel

ARGS = {
    "image": lab_image
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — 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
linear_color_transfer Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("d30d467e1fd0d091")
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 (4, 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 small sample images with distinct dimensions for clarity
# Images expected to be (width, height, channels) in float [0, 1]
np.random.seed(0)

# Target image: 4x3 pixels, 3 channels (RGB)
target_img = np.random.rand(4, 3, 3).astype(np.float32)
target_img = np.clip(target_img, 0, 1)

# Source image: 5x3 pixels, 3 channels (RGB) - different size to show distinction
source_img = np.random.rand(5, 3, 3).astype(np.float32)
source_img = np.clip(source_img, 0, 1)

ARGS = {
    'target_img': target_img,
    'source_img': source_img,
    'mode': 'pca',
    'eps': 1e-05
}
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 float, shape (4, 3, 3) — shape (4, 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
mask_postprocess Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("a7f3d57c95a1c2f0")
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 2, dtype family int, shape (64, 48) 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 cv2

# Create a sample mask as a numpy array
# The mask should be a 2D array suitable for cv2 operations
# Using uint8 type as expected by cv2.erode and cv2.dilate
mask = np.random.randint(0, 256, size=(64, 48), dtype=np.uint8)

ARGS = {
    "mask": mask
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family int, shape (64, 48) — shape (64, 48) 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
random_deform Ran
ran; our contract draft was wrong, not the code
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("b97c47134fc596c6")
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 2, 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
import cv2
from skimage.transform import PiecewiseAffineTransform, warp

# Create a small test mask with distinct dimensions (4x3 for clarity)
mask = np.random.randint(0, 255, size=(4, 3), dtype=np.uint8)

# Define grid parameters
nrows = 2
ncols = 2
mean = 0
std = 10

# Construct ARGS dictionary with all parameters
ARGS = {
    'mask': mask,
    'nrows': nrows,
    'ncols': ncols,
    'mean': mean,
    'std': std
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy, skimage — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float — no output shape is stated: observe_signature.py could not observe it (call:QhullError)
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
reinhard_color_transfer Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("7d79e91f53c5b500")
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 (4, 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
import cv2

# Create small test images in BGR format
# Using distinct dimensions: 4x3 for target, 5x3 for source
np.random.seed(0)
target_image = np.random.randint(0, 256, size=(4, 3, 3), dtype=np.uint8)
source_image = np.random.randint(0, 256, size=(5, 3, 3), dtype=np.uint8)

# Create optional mask arrays with same spatial dimensions as images
# but normalized to [0, 1] for masking purposes
target_mask = np.ones((4, 3, 3), dtype=np.float32)
source_mask = np.ones((5, 3, 3), dtype=np.float32)

ARGS = {
    'target': target_image,
    'source': source_image,
    'clip': True,
    'preserve_paper': False,
    'source_mask': source_mask,
    'target_mask': target_mask,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 (4, 3, 3) — shape (4, 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
remove_eyes Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("f976c91ae7d4e300")
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 2, dtype family float, shape (100, 80) 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 math
import numpy as np
import cv2
from scipy.ndimage import binary_erosion, binary_dilation

# Create a small test image (height=100, width=80, channels=3)
image = np.random.randint(0, 255, (100, 80, 3), dtype=np.uint8)

# Create landmarks as a list of (x, y) tuples
# Need at least 9 landmarks for indices 0-2 and 5-9
landmarks = [
    (10, 20),   # 0: brow left
    (15, 25),   # 1: brow right
    (30, 40),   # 2: unused in this function but included for completeness
    (35, 45),   # 3: unused
    (40, 50),   # 4: unused
    (45, 55),   # 5: left eye point 1
    (50, 60),   # 6: left eye point 2
    (55, 65),   # 7: right eye point 1
    (60, 70),   # 8: right eye point 2
]

# Create opt parameter - can be 'l', 'r', or 'b'
opt = 'l'

ARGS = {
    'image': image,
    'landmarks': landmarks,
    'opt': opt
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (100, 80) — shape (100, 80) 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
remove_mouth Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("773f480524c30a0b")
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 2, dtype family float, shape (10, 12) 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 math
import numpy as np
import cv2
from scipy.ndimage import binary_erosion, binary_dilation

# Create a small test image with 3 channels (BGR format for cv2)
# Using shape (10, 12, 3) to have distinct dimensions
image = np.random.randint(0, 255, size=(10, 12, 3), dtype=np.uint8)

# Create landmarks with 5 points (as the function accesses indices 3:5)
# Each landmark is (x, y) coordinates
landmarks = [
    (2, 2),    # 0
    (3, 3),    # 1
    (4, 4),    # 2
    (2, 5),    # 3 - mouth point 1
    (6, 5),    # 4 - mouth point 2
]

ARGS = {
    "image": image,
    "landmarks": landmarks,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (10, 12) — shape (10, 12) 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
remove_nose Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("17b7d389e847463d")
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 2, dtype family float, shape (64, 64) 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 cv2

# Create a small RGB image
image = np.random.randint(0, 256, size=(64, 64, 3), dtype=np.uint8)

# Create landmarks: 3 points (2D coordinates)
# landmarks[:2] are used for eyes (x1, y1) and (x2, y2)
# landmarks[2] is used for nose (x3, y3)
landmarks = np.array([
    (10, 20),   # left eye
    (50, 25),   # right eye
    (30, 40)    # nose
], dtype=np.int32)

ARGS = {
    "image": image,
    "landmarks": landmarks
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (64, 64) — shape (64, 64) 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
umeyama Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("e466d89b39b88402")
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 2, dtype family float, shape (4, 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 numpy as np

np.random.seed(0)

# Create source and destination point clouds
# M=5 points, N=3 dimensions
src = np.random.randn(5, 3).astype(np.float64)
dst = np.random.randn(5, 3).astype(np.float64)

# estimate_scale parameter
estimate_scale = True

ARGS = {
    "src": src,
    "dst": dst,
    "estimate_scale": estimate_scale,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (4, 4) — shape (4, 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
RPModel Not yet run
unverified (harvested, no recorded run)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("b7583eca3a5dea96")
How it ran

no invocation recorded.

TransferModel Not yet run
unverified (harvested, no recorded run)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("2df4fb5ac63e85f0")
How it ran

no invocation recorded.

align_eyes Not yet run
unverified (harvested, no recorded run)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("5d568a8c6b4e3c8f")
How it ran

no invocation recorded.

blend_fake_to_real Not yet run
unverified (harvested, no recorded run)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("8ec3fc1b4f8772c6")
How it ran

no invocation recorded.

color_transfer Not yet run
unverified (harvested, no recorded run)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("11b54e2c077cce24")
How it ran

no invocation recorded.

get_5_keypoint Not yet run
unverified (harvested, no recorded run)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("0cac967602788bdb")
How it ran

no invocation recorded.

get_align_mat_new Not yet run
unverified (harvested, no recorded run)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("3f8cf674f1451c2a")
How it ran

no invocation recorded.

synthesizer Not yet run
unverified (harvested, no recorded run)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("e49f897dc62e508e")
How it ran

no invocation recorded.

xception Not yet run
unverified (harvested, no recorded run)
liangchen527/SLADD/src/networks/synthesizer.py
code served (permissive licence) · get_code("6cc4b069300d3ce0")
How it ran

no invocation recorded.

Also present in other papers' code

Of 36 samples lifted from this paper's linked repositories, 36 carry a function identity and 0 do not. Of the 36 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 35 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
umeyama
brick of this paper's assembly: helper
1 other paper
1 dated earlier
1903.07291 (2019-03)
Semantic Image Synthesis with Spatially-Adaptive Normalization

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

Recent studies in deepfake detection have yielded promising results when the training and testing face forgeries are from the same dataset. However, the problem remains challenging when one tries to generalize the detector to forgeries created by unseen methods in the training dataset. This work addresses the generalizable deepfake detection from a simple principle: a generalizable representation should be sensitive to diverse types of forgeries. Following this principle, we propose to enrich the "diversity" of forgeries by synthesizing augmented forgeries with a pool of forgery configurations and strengthen the "sensitivity" to the forgeries by enforcing the model to predict the forgery configurations. To effectively explore the large forgery augmentation space, we further propose to use the adversarial training strategy to dynamically synthesize the most challenging forgeries to the current model. Through extensive experiments, we show that the proposed strategies are surprisingly effective (see Figure 1), and they could achieve superior performance than the current state-of-the-art methods. Code is available at https://github.com/liangchen527/SLADD.

For agents

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

get_harvested_code_for_paper("2203.12208")
get_code_for_paper("2203.12208")
have("2203.12208")

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

curl https://syntology.ai/api/ran/2203.12208.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.12208.svg)](https://syntology.ai/paper/2203.12208)

Connect an agent — have() is free.