SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.03625 · 2026

Multi-Objective Optimization for Synthetic-to-Real Style Transfer

Thomas Oberlin, Estelle Chigot, Manon Huguenin, Dennis Wilson

arXiv · PDF

Code that ran

10 of 13 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; 5 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 5 failures of our instrument; 3 no recorded run), measured 2026-11-22; 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/2602.03625.json · badge: /api/ran/2602.03625.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
echigot/MOOSS canonical (classifier label)
classifier basis: census
10 ran · 3 no recorded run
FunctionStatusWhere it lives
adain Ran
ran; our contract draft was wrong, not the code
echigot/MOOSS/mooss/adain.py
code served (permissive licence) · get_code("bfeae26f4ed21855")
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 (4, 3, 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create two feature tensors with shape (C, H, W) - 3D tensors
# C=4 channels, H=3 height, W=5 width to have distinct dimensions
content_feat = torch.randn(4, 3, 5)
style_feat = torch.randn(4, 3, 5)

ARGS = {
    'content_feat': content_feat,
    'style_feat': style_feat,
}
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 (4, 3, 5) — shape (4, 3, 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
apply_augmentation Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
echigot/MOOSS/mooss/augmentation.py
code served (permissive licence) · get_code("823b21ee472f337f")
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 Image. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
from PIL import Image, ImageEnhance, ImageFilter
import numpy as np

torch.manual_seed(0)

# Create minimal PIL Images
def create_test_image(size=(64, 48), color=(100, 150, 200)):
    return Image.new('RGB', size, color)

def create_test_label(size=(64, 48)):
    # Create a simple segmentation label image
    arr = np.random.randint(0, 20, size=(*size, 1), dtype=np.uint8)
    return Image.fromarray(arr.squeeze(), mode='L')

# Mock functions that are called in the augmentation pipeline
def generate_canny_image(image):
    """Mock canny edge detection"""
    return image.convert('L')

def convert_label_to_ade(label):
    """Mock label conversion"""
    return label

def load_controlnet_pipeline():
    """Mock pipeline loader - return a mock pipe object"""
    class MockPipe:
        def __init__(self):
            self.vae = MockVAE()
            self.image_processor = MockImageProcessor()
    return MockPipe()

def run_inference(pipe, reference, ade_label, canny_image, style_name=''):
    """Mock inference"""
    return [reference]

def normalize_with_reference(image, reference):
    """Mock normalization"""
    return image

def adain(content_latents, style_latents):
    """Mock AdaIN"""
    return content_latents

class MockVAE:
    def __init__(self):
        self.config = type('obj', (object,), {'scaling_factor': 0.18215})()
    
    def encode(self, x):
        class LatentDist:
            def mode(self):
                return torch.randn(1, 4, 8, 6)
        return type('obj', (object,), {'latent_dist': LatentDist()})()
    
    def decode(self, x, return_dict=False):
        decoded = torch.randn(1, 3, 64, 48)
        if return_dict:
            return {'sample': decoded}
        return (decoded,)

class MockImageProcessor:
    def preprocess(self, image, width=1024, height=512):
        return torch.randn(1, 3, height, width)
    
    def postprocess(self, latents, output_type='pil', do_denormalize=None):
        img = Image.new('RGB', (64, 48), (100, 100, 100))
        return [img]

# Construct ARGS
ARGS = {
    'image': create_test_image(size=(64, 48), color=(100, 150, 200)),
    'label': create_test_label(size=(64, 48)),
    'augmentation_name': 'blur',
    'reference': create_test_image(size=(64, 48), color=(200, 100, 150)),
    'params': {'radius': 2.0, 'alpha': 0.5, 'factor': 1.5},
    'pipe': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
calc_mean_std Ran
ran; our contract draft was wrong, not the code
echigot/MOOSS/mooss/adain.py
code served (permissive licence) · get_code("7ee9dfbb0bd00480")
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 4, 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)

# Test case 1: 3D tensor without mask
feat_3d = torch.randn(4, 3, 5)  # C=4, H=3, W=5

# Test case 2: 3D tensor with mask
mask_3d = torch.ones(3, 5)
mask_3d[0, 0] = 0  # Mark one position as masked out

# Test case 3: 2D tensor without mask
feat_2d = torch.randn(4, 15)  # C=4, flattened spatial

# Test case 4: 2D tensor with mask
mask_2d = torch.ones(15)
mask_2d[0] = 0

# Use the 3D case as primary test
ARGS = {
    "feat": feat_3d,
    "eps": 1e-05,
    "mask": None
}
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 4, 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
calc_mean_std_2d Ran
ran; our contract draft was wrong, not the code
echigot/MOOSS/mooss/adain.py
code served (permissive licence) · get_code("dcbb6d9b7d233a4e")
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 3, 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

torch.manual_seed(0)

# Create a 2D feature tensor with shape (C, N) where C=4 channels, N=6 elements
feat = torch.randn(4, 6)

# Create a mask tensor with shape (6,) where 1 indicates valid elements
mask = torch.tensor([1, 1, 1, 0, 1, 1])

ARGS = {
    "feat": feat,
    "eps": 1e-05,
    "mask": mask,
}
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 3, 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
city_to_ade20k Ran
ran, honoured the contract we drafted
echigot/MOOSS/mooss/gta2ade.py
code served (permissive licence) · get_code("b6e7991413068440")
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 Image. 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 numpy as np
from PIL import Image

# Create a small 2D label array with shape (4, 3) representing cityscapes labels
# Values should be in range [0, 19] since there are 20 cityscapes classes
city_label = np.array([[0, 1, 2], [3, 4, 5], [6, 7, 8], [9, 10, 11]], dtype=np.uint8)

ARGS = {
    'city_label': city_label
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
compute_embeddings_for_dir Ran
ran, honoured the contract we drafted
echigot/MOOSS/mooss/cmmd.py
code served (permissive licence) · get_code("cddf1462a8515cca")
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, 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 numpy as np
from pathlib import Path
from torch.utils.data import DataLoader
import tempfile
import torch
from PIL import Image
import os

torch.manual_seed(0)

# Create a temporary directory with a few test images
temp_dir = tempfile.mkdtemp()

# Create 3 small test images (4x4 RGB images)
for i in range(3):
    img_array = np.random.randint(0, 256, (4, 4, 3), dtype=np.uint8)
    img = Image.fromarray(img_array, mode='RGB')
    img.save(os.path.join(temp_dir, f'test_image_{i}.png'))

# Define a minimal embedding model class that satisfies the interface
class MockEmbeddingModel:
    def __init__(self, input_size, embed_dim):
        self.input_image_size = input_size
        self.embed_dim = embed_dim
    
    def embed(self, image_batch):
        """
        Args:
            image_batch: numpy array of shape (batch_size, height, width, channels)
                        with values in [0, 1]
        
        Returns:
            embeddings: numpy array of shape (batch_size, embed_dim)
        """
        batch_size = image_batch.shape[0]
        return np.random.randn(batch_size, self.embed_dim).astype('float32')

embedding_model = MockEmbeddingModel(input_size=4, embed_dim=8)

ARGS = {
    'img_dir': temp_dir,
    'embedding_model': embedding_model,
    'batch_size': 2,
    'max_count': 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch, tqdm — 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, 8) — shape (2, 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
generate_canny_image Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
echigot/MOOSS/mooss/controlnet_utils.py
code served (permissive licence) · get_code("5fa1659e1f3c0b9a")
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 Image. 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 cv2
import numpy as np
from PIL import Image

# Create a small test image
# The function expects input_image to be convertible to numpy array
# and have shape that cv2.Canny can process
test_image_array = np.random.randint(0, 256, (64, 48, 3), dtype=np.uint8)
test_image = Image.fromarray(test_image_array)

ARGS = {
    "input_image": test_image
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, cv2, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
load_images_as_tensors Ran
ran, honoured the contract we drafted
echigot/MOOSS/mooss/metrics.py
code served (permissive licence) · get_code("9bd6d13571397d5d")
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, 3, 512, 1024) 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
from pathlib import Path
from PIL import Image
import tempfile
import os

torch.manual_seed(0)

# Create a temporary directory with some test PNG images
temp_dir = tempfile.mkdtemp()
image_dir = Path(temp_dir)

# Create 2 small test images (1024x512 as expected by the function)
for i in range(2):
    # Create a small RGB image
    img_array = np.random.randint(0, 256, (512, 1024, 3), dtype=np.uint8)
    img = Image.fromarray(img_array, 'RGB')
    img_path = image_dir / f"test_image_{i}.png"
    img.save(img_path)

ARGS = {
    'image_dir': image_dir,
    'max_images': 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, 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 4, dtype family float, shape (2, 3, 512, 1024) — shape (2, 3, 512, 1024) 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
load_specific_images_as_tensors Ran
ran, honoured the contract we drafted
echigot/MOOSS/mooss/metrics.py
code served (permissive licence) · get_code("99ede2f82351ddf4")
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, 3, 512, 1024) 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
from PIL import Image
import tempfile
import os

torch.manual_seed(0)

# Create temporary directory for test images
temp_dir = tempfile.mkdtemp()

# Create small test images and save them
image_paths = []
for i in range(2):
    # Create a small RGB image (4x3 pixels)
    img_array = np.random.randint(0, 256, (3, 4, 3), dtype=np.uint8)
    img = Image.fromarray(img_array, mode='RGB')
    img_path = os.path.join(temp_dir, f'test_image_{i}.png')
    img.save(img_path)
    image_paths.append(img_path)

ARGS = {
    'image_paths': image_paths
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, 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 4, dtype family float, shape (2, 3, 512, 1024) — shape (2, 3, 512, 1024) 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
mmd Ran
ran, honoured the contract we drafted
echigot/MOOSS/mooss/cmmd.py
code served (permissive licence) · get_code("99ac6ca026cab870")
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 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 torch
import numpy as np

torch.manual_seed(0)

# Create small numpy arrays for x and y
# Shape: (n, embedding_dim) where n=2, embedding_dim=4
x = np.random.randn(2, 4).astype(np.float32)
y = np.random.randn(2, 4).astype(np.float32)

ARGS = {
    "x": x,
    "y": y,
}
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 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
convert_label_to_ade Not yet run
unverified (harvested, no recorded run)
echigot/MOOSS/mooss/controlnet_utils.py
code served (permissive licence) · get_code("15b9df8b368d8e40")
How it ran

no invocation recorded.

extract_hyperparameters Not yet run
unverified (harvested, no recorded run)
echigot/MOOSS/mooss/augmentation.py
code served (permissive licence) · get_code("07f4c924d4f6edaa")
How it ran

no invocation recorded.

gta_to_ade20k Not yet run
unverified (harvested, no recorded run)
echigot/MOOSS/mooss/gta2ade.py
code served (permissive licence) · get_code("95d5af05e9d52b9f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 0 later) and 10 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
calc_mean_std 2 other papers
2 dated earlier
2401.02126 (2024-01)
Unified Diffusion-Based Rigid and Non-Rigid Editing with Text and Image Guidance
2407.17850 (2024-07)
FlexiEdit: Frequency-Aware Latent Refinement for Enhanced Non-Rigid Editing
calc_mean_std_2d 2 other papers
2 dated earlier
2401.02126 (2024-01)
Unified Diffusion-Based Rigid and Non-Rigid Editing with Text and Image Guidance
2407.17850 (2024-07)
FlexiEdit: Frequency-Aware Latent Refinement for Enhanced Non-Rigid Editing
adain 1 other paper
1 dated earlier
2308.03202 (2023-08)
Source-free Domain Adaptive Human Pose Estimation

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

Semantic segmentation networks require large amounts of pixel-level annotated data, which are costly to obtain for real-world images. Computer graphics engines can generate synthetic images alongside their ground-truth annotations. However, models trained on such images can perform poorly on real images due to the domain gap between real and synthetic images. Style transfer methods can reduce this difference by applying a realistic style to synthetic images. Choosing effective data transformations and their sequence is difficult due to the large combinatorial search space of style transfer operators. Using multi-objective genetic algorithms, we optimize pipelines to balance structural coherence and style similarity to target domains. We study the use of pairedimage metrics on individual image samples during evolution to enable rapid pipeline evaluation, as opposed to standard distributional metrics that require the generation of many images. After optimization, we evaluate the resulting Pareto front using distributional metrics and segmentation performance. We apply this approach to standard datasets in synthetic-to-real domain adaptation: from the video game GTA5 to real image datasets Cityscapes and ACDC, focusing on adverse conditions. Results demonstrate that evolutionary algorithms can propose diverse augmentation pipelines adapted to different objectives. The contribution of this work is the formulation of style transfer as a sequencing problem suitable for evolutionary optimization and the study of efficient metrics that enable feasible search in this space. The source code is available at: https://github.com/echigot/MOOSS.

For agents

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

get_harvested_code_for_paper("2602.03625")
get_code_for_paper("2602.03625")
have("2602.03625")

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

curl https://syntology.ai/api/ran/2602.03625.json

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

[![Syntology run record](https://syntology.ai/api/ran/2602.03625.svg)](https://syntology.ai/paper/2602.03625)

Connect an agent — have() is free.