SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1803.01837 · 2018

ST-GAN: Spatial Transformer Generative Adversarial Networks for Image Compositing

arXiv · PDF · Open in the Atlas

Code that ran

11 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; 0 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 2 failures of our instrument; 2 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/1803.01837.json · badge: /api/ran/1803.01837.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
chenhsuanlin/spatial-transformer-GAN canonical (classifier label)
classifier basis: census
9 ran · 1 no recorded run
lix3972/https-github.com-chenhsuanlin-spatial-transformer-GAN community (classifier label)
classifier basis: plan_row
2 ran · 1 no recorded run
FunctionStatusWhere it lives
compose Ran
ran on a synthesized input
chenhsuanlin/spatial-transformer-GAN/glasses/warp.py
code served (permissive licence) · get_code("d333acbd22f1fd12")
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
import numpy as np

torch.manual_seed(0)

# The function compose simply adds two arguments together: p + dp
# So opt, p, and dp need to support addition
# opt is not used in the function body
# p and dp should be compatible for addition

# Create simple tensor arguments
p = torch.randn(2, 4)
dp = torch.randn(2, 4)
opt = {}

ARGS = {
    "opt": opt,
    "p": p,
    "dp": dp
}
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 (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
composite Ran
ran on a synthesized input
chenhsuanlin/spatial-transformer-GAN/glasses/graph.py
code served (permissive licence) · get_code("b1c0d809510d7dee")
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 4, dtype family float, shape (2, 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 tensorflow as tf
import numpy as np

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

# Create tiny test tensors
# imageFG: [batch, height, width, channels] where channels = 4 (RGB + alpha mask)
# Shape: [2, 4, 3, 4] - batch=2, height=4, width=3, channels=4
imageFG = tf.constant(np.random.rand(2, 4, 3, 4).astype(np.float32))

# imageBG: [batch, height, width, channels] where channels = 3 (RGB only)
# Shape: [2, 4, 3, 3] - batch=2, height=4, width=3, channels=3
imageBG = tf.constant(np.random.rand(2, 4, 3, 3).astype(np.float32))

# opt: optimizer or configuration object
# The function doesn't use opt in the body, so a simple dict works
opt = {}

ARGS = {
    'opt': opt,
    'imageBG': imageBG,
    'imageFG': imageFG
}
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 4, dtype family float, shape (2, 4, 3, 3) — shape (2, 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
fit Ran
ran on a synthesized input
chenhsuanlin/spatial-transformer-GAN/glasses/warp.py
code served (permissive licence) · get_code("5dfbbadd25cd659e")
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 (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

# The fit function expects two numpy arrays:
# - Xsrc: source points, shape (N, 2) where N is number of points
# - Xdst: destination points, shape (N, 2)
# The function computes a 2D affine transformation matrix

# Create small sample point sets
# Using 4 points for a small but sufficient example
Xsrc = np.array([
    [0.0, 0.0],
    [1.0, 0.0],
    [0.0, 1.0],
    [1.0, 1.0]
], dtype=np.float32)

# Destination points (slightly transformed version)
Xdst = np.array([
    [0.5, 0.5],
    [1.5, 0.3],
    [0.2, 1.5],
    [1.3, 1.2]
], dtype=np.float32)

ARGS = {
    "Xsrc": Xsrc,
    "Xdst": Xdst
}
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 2, dtype family float, shape (3, 3) — shape (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
imread Ran
ran on a synthesized input
chenhsuanlin/spatial-transformer-GAN/glasses/util.py
code served (permissive licence) · get_code("f027fae9fd59cbd6")
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 imageio
import numpy as np
from io import BytesIO
from PIL import Image

# Create a small valid image file in memory
img_array = np.random.randint(0, 256, (4, 3, 3), dtype=np.uint8)
img = Image.fromarray(img_array)
img_bytes = BytesIO()
img.save(img_bytes, format='PNG')
img_bytes.seek(0)

# Save to a temporary file-like path we can reference
# Since we can't do real file I/O, we'll create a mock that imageio can read
import tempfile
import os

temp_dir = tempfile.gettempdir()
temp_file = os.path.join(temp_dir, 'test_image.png')

# Create actual test image
test_img_array = np.random.randint(0, 256, (4, 3, 3), dtype=np.uint8)
test_img = Image.fromarray(test_img_array)
test_img.save(temp_file)

ARGS = {
    'fname': temp_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, imageio, 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
inverse Ran
ran on a synthesized input
chenhsuanlin/spatial-transformer-GAN/glasses/warp.py
code served (permissive licence) · get_code("e2db836dae92016c")
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
import numpy as np

torch.manual_seed(0)

# The function takes two parameters: opt and p
# Looking at the function body: return -p
# This means p needs to support negation (unary minus operator)
# p is most likely a tensor or numpy array

# opt appears to be unused in the function body, so it can be any object
# Let's make it a simple dict (common for optimizer/options)

opt = {
    "lr": 0.01,
    "momentum": 0.9
}

# p should be a tensor that supports negation
p = torch.randn(2, 4)

ARGS = {
    "opt": opt,
    "p": p
}
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 (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
load Ran
ran on a synthesized input
chenhsuanlin/spatial-transformer-GAN/glasses/data.py
code served (permissive licence) · get_code("bc55ffac0f03d4f5")
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 dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import tempfile
import os

# Create a temporary directory to simulate the dataset structure
temp_dir = tempfile.mkdtemp()
dataset_path = os.path.join(temp_dir, 'dataset')
os.makedirs(dataset_path, exist_ok=True)

# Create mock numpy arrays for the dataset files
# Training images: 10 samples, 64x64x3
train_images = np.random.randint(0, 256, (10, 64, 64, 3), dtype=np.uint8)
# Training attributes: 10 samples, 40 attributes (glasses is at index 15)
train_attributes = np.random.randint(0, 2, (10, 40), dtype=np.int32)
# Test images: 5 samples, 64x64x3
test_images = np.random.randint(0, 256, (5, 64, 64, 3), dtype=np.uint8)
# Test attributes: 5 samples, 40 attributes
test_attributes = np.random.randint(0, 2, (5, 40), dtype=np.int32)
# Glasses template: a small image, e.g., 32x32x3
glasses = np.random.randint(0, 256, (32, 32, 3), dtype=np.uint8)

# Save the mock data
np.save(os.path.join(dataset_path, 'image_train.npy'), train_images)
np.save(os.path.join(dataset_path, 'attribute_train.npy'), train_attributes)
np.save(os.path.join(dataset_path, 'image_test.npy'), test_images)
np.save(os.path.join(dataset_path, 'attribute_test.npy'), test_attributes)
np.save(os.path.join(dataset_path, 'glasses.npy'), glasses)

# Change to the temp directory so the relative path 'dataset' works
original_cwd = os.getcwd()
os.chdir(temp_dir)

# Define opt as a simple dict or object
opt = {}

ARGS = {
    'opt': opt,
    'test': False
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_homo Ran
ran on a synthesized input
chenhsuanlin/spatial-transformer-GAN/indoor/data.py
code served (permissive licence) · get_code("2a185557c2573563")
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 dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import tempfile
import os

# Create temporary directory structure and dummy .npy files
tmpdir = tempfile.mkdtemp()
train_dir = os.path.join(tmpdir, 'dataset', 'train')
test_dir = os.path.join(tmpdir, 'dataset', 'test')
os.makedirs(train_dir, exist_ok=True)
os.makedirs(test_dir, exist_ok=True)

# Create dummy numpy files for both train and test
for directory in [train_dir, test_dir]:
    np.save(os.path.join(directory, 'disocclude.npy'), np.zeros((2, 3)))
    np.save(os.path.join(directory, 'removed.npy'), np.zeros((2, 4)))
    np.save(os.path.join(directory, 'indiv_mask.npy'), np.zeros((2, 5)))

# Change to the temp directory so relative paths work
original_dir = os.getcwd()
os.chdir(tmpdir)

class MockOpt:
    pass

opt = MockOpt()

ARGS = {
    'opt': opt,
    'test': False
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
makeBatch Ran
ran on a synthesized input
chenhsuanlin/spatial-transformer-GAN/glasses/data.py
code served (permissive licence) · get_code("ad896d56681f4a3d")
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 dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

class OptConfig:
    def __init__(self, batchSize):
        self.batchSize = batchSize

opt = OptConfig(batchSize=2)

data = {
    'image0': np.random.randint(0, 256, size=(5, 4, 3, 3), dtype=np.uint8),
    'image1': np.random.randint(0, 256, size=(6, 4, 3, 3), dtype=np.uint8),
    'glasses': np.random.randint(0, 256, size=(7, 4, 3, 3), dtype=np.uint8),
}

imageBGfakeData = 'placeholder_bg'
imageRealData = 'placeholder_real'
imageFGfake = 'placeholder_fg'
PH = [imageBGfakeData, imageRealData, imageFGfake]

ARGS = {
    'opt': opt,
    'data': data,
    'PH': PH,
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
makeBatchEval Ran
ran on a synthesized input
chenhsuanlin/spatial-transformer-GAN/glasses/data.py
code served (permissive licence) · get_code("247b153b64fa322c")
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 dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

class OptConfig:
    def __init__(self, batchSize):
        self.batchSize = batchSize

opt = OptConfig(batchSize=2)

testImage = np.random.randint(0, 256, size=(1, 4, 5, 3), dtype=np.uint8)

glasses = np.random.randint(0, 256, size=(2, 4, 5, 3), dtype=np.uint8)

imageBG = "imageBG"
imageFG = "imageFG"
PH = [imageBG, imageFG]

ARGS = {
    "opt": opt,
    "testImage": testImage,
    "glasses": glasses,
    "PH": PH
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
toGreen Ran
ran; our contract draft was wrong, not the code
lix3972/https-github.com-chenhsuanlin-spatial-transformer-GAN/glasses/util.py
code served (permissive licence) · get_code("045caed495c3be39")
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 str. 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 termcolor

ARGS = {
    'content': 'Hello, World!'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
termcolor — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
toRed Ran
ran; our contract draft was wrong, not the code
lix3972/https-github.com-chenhsuanlin-spatial-transformer-GAN/glasses/util.py
code served (permissive licence) · get_code("23debb41fe868b57")
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 str. 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 termcolor

def toRed(content):
    return termcolor.colored(content, 'red', attrs=['bold'])

# The function takes a single parameter 'content' which is passed to termcolor.colored()
# termcolor.colored() expects a string as the first argument
# We'll pass a simple string

ARGS = {
    'content': 'Hello World'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
termcolor — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
discriminator Not yet run
unverified (harvested, no recorded run)
chenhsuanlin/spatial-transformer-GAN/glasses/graph.py
code served (permissive licence) · get_code("3c02b07781bd054e")
How it ran

no invocation recorded.

imread Not yet run
unverified (harvested, no recorded run)
lix3972/https-github.com-chenhsuanlin-spatial-transformer-GAN/glasses/util.py
code served (permissive licence) · get_code("4497ff1ceb01b614")
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, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 2 later) and 9 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
toGreen 2 other papers
1 dated earlier, 1 later
1612.03897 (2016-12)
Inverse Compositional Spatial Transformer Networks
2403.02611 (2024-03)
A Unified Framework for Microscopy Defocus Deblur with Multi-Pyramid Transformer and Contrastive Learning
toRed 2 other papers
1 dated earlier, 1 later
1612.03897 (2016-12)
Inverse Compositional Spatial Transformer Networks
2403.02611 (2024-03)
A Unified Framework for Microscopy Defocus Deblur with Multi-Pyramid Transformer and Contrastive Learning
fit 1 other paper
1 dated earlier
1612.03897 (2016-12)
Inverse Compositional Spatial Transformer Networks
imread 1 other paper
1 dated earlier
1612.03897 (2016-12)
Inverse Compositional Spatial Transformer Networks

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

We address the problem of finding realistic geometric corrections to a foreground object such that it appears natural when composited into a background image. To achieve this, we propose a novel Generative Adversarial Network (GAN) architecture that utilizes Spatial Transformer Networks (STNs) as the generator, which we call Spatial Transformer GANs (ST-GANs). ST-GANs seek image realism by operating in the geometric warp parameter space. In particular, we exploit an iterative STN warping scheme and propose a sequential training strategy that achieves better results compared to naive training of a single generator. One of the key advantages of ST-GAN is its applicability to high-resolution images indirectly since the predicted warp parameters are transferable between reference frames. We demonstrate our approach in two applications: (1) visualizing how indoor furniture (e.g. from product images) might be perceived in a room, (2) hallucinating how accessories like glasses would look when matched with real portraits.

For agents

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

get_harvested_code_for_paper("1803.01837")
get_code_for_paper("1803.01837")
have("1803.01837")

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

curl https://syntology.ai/api/ran/1803.01837.json

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

[![Syntology run record](https://syntology.ai/api/ran/1803.01837.svg)](https://syntology.ai/paper/1803.01837)

Connect an agent — have() is free.