SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1504.06852 · 2015

FlowNet: Learning Optical Flow with Convolutional Networks

arXiv · PDF · Open in the Atlas

Code that ran

13 of 20 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 1 violated it and 12 have no contract verdict; 0 failures of our instrument; 7 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/1504.06852.json · badge: /api/ran/1504.06852.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
McDo/LightFlowPytorch community (classifier label)
classifier basis not recorded on the link
12 ran · 5 no recorded run
ClementPinard/Pytorch-Correlation-extension community (classifier label)
classifier basis not recorded on the link
1 ran · 0 no recorded run
ClementPinard/FlowNetPytorch reimplementation (classifier label)
classifier basis: classifier_2025
0 ran · 2 no recorded run
FunctionStatusWhere it lives
EPE Ran
ran on a synthesized input
McDo/LightFlowPytorch/loss.py
code served (permissive licence) · get_code("b77a7cb2a29a19c7")
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

torch.manual_seed(0)

# Create small flow tensors
# input_flow and target_flow should have shape (batch_size, 2, height, width)
# or (batch_size, 2) for this function since it uses dim=1
batch_size = 2
channels = 2

input_flow = torch.randn(batch_size, channels)
target_flow = torch.randn(batch_size, channels)

ARGS = {
    "input_flow": input_flow,
    "target_flow": target_flow,
    "sparse": False,
    "mean": True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 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
L1Loss Ran
ran on a synthesized input
McDo/LightFlowPytorch/loss.py
code served (permissive licence) · get_code("e58f02bc54cf9db9")
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 torch.nn.functional as F

torch.manual_seed(0)

# Create small tensors for testing
# output: (batch=2, channels=4, height=8, width=8)
output = torch.randn(2, 4, 8, 8)

# target: (batch=2, channels=4, height=16, width=16) - different spatial dims
target = torch.randn(2, 4, 16, 16)

# sparse: boolean flag
sparse = True

ARGS = {
    "output": output,
    "target": target,
    "sparse": sparse
}
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 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
L2Loss Ran
ran on a synthesized input
McDo/LightFlowPytorch/loss.py
code served (permissive licence) · get_code("1869efdefff249d5")
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 torch.nn.functional as F

torch.manual_seed(0)

# Create small tensors for the function
batch_size = 2
channels = 4
height = 8
width = 6

# output tensor: (batch, channels, height, width)
output = torch.randn(batch_size, channels, height, width)

# target tensor: (batch, channels, height, width) - typically larger or different size
target = torch.randn(batch_size, channels, height * 2, width * 2)

# sparse flag: boolean
sparse = True

ARGS = {
    "output": output,
    "target": target,
    "sparse": sparse
}
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 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
adjust_learning_rate Ran
ran on a synthesized input
McDo/LightFlowPytorch/utils/util.py
code served (permissive licence) · get_code("97509470e0c1680e")
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 torch
import torch.nn as nn
import torch.optim as optim

torch.manual_seed(0)

# Create a minimal model with parameters to optimize
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

model = SimpleModel()
optimizer = optim.SGD(model.parameters(), lr=0.1)

# Create args object with required attributes
class Args:
    def __init__(self):
        self.lr = 0.1

args = Args()
epoch = 5
epoch_size = 100

ARGS = {
    "args": args,
    "optimizer": optimizer,
    "epoch": epoch,
    "epoch_size": epoch_size
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
compute_color Ran
ran on a synthesized input
McDo/LightFlowPytorch/utils/flowlib.py
code served (permissive licence) · get_code("ab41dcb0fc2e9217")
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, 5, 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 function compute_color takes two parameters:
# u: optical flow horizontal map (2D numpy array)
# v: optical flow vertical map (2D numpy array)

# Create small sample optical flow arrays
h, w = 4, 5  # height and width
u = np.random.randn(h, w).astype(np.float32) * 0.5
v = np.random.randn(h, w).astype(np.float32) * 0.5

ARGS = {
    'u': u,
    'v': v
}
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, 5, 3) — shape (4, 5, 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
conv_dw Ran
ran on a synthesized input
McDo/LightFlowPytorch/models/FlowNetL.py
code served (permissive licence) · get_code("fd0432127db8b8c0")
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 Sequential. 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 = {
    "in_planes": 3,
    "out_planes": 16,
    "kernel_size": 3,
    "stride": 2,
}
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 Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential output the recorded ndim 1 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
flow_to_image Ran
ran on a synthesized input
McDo/LightFlowPytorch/utils/flowlib.py
code served (permissive licence) · get_code("11b939d4bfc443d5")
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 a small optical flow array
# flow has shape (height, width, 2) where the last dimension is (u, v) components
height, width = 4, 3
flow = np.random.randn(height, width, 2).astype(np.float32) * 5.0

ARGS = {
    "flow": flow
}
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
get_grads Ran
ran on a synthesized input
ClementPinard/Pytorch-Correlation-extension/check.py
code served (permissive licence) · get_code("19ee68e6a7f1011b")
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. 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)

# Create a simple model with parameters that will have gradients
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

model = SimpleModel()

# Create input and target tensors
x = torch.randn(2, 4)
y = torch.randn(2, 2)

# Forward pass
output = model(x)

# Compute loss and backward
loss_fn = nn.MSELoss()
loss = loss_fn(output, y)
loss.backward()

# Collect variables (model parameters) that have gradients
variables = [param for param in model.parameters() if param.grad is not None]

ARGS = {
    "variables": variables
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list output has no array shape of its own; none is 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
make_dataset Ran
ran on a synthesized input
McDo/LightFlowPytorch/datasets/KITTI.py
code served (permissive licence) · get_code("97658411b60eb6fa")
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. 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 os
import tempfile
import shutil

# Create a temporary directory structure that mimics KITTI dataset layout
temp_dir = tempfile.mkdtemp()

# Create subdirectories
flow_occ_dir = os.path.join(temp_dir, 'flow_occ')
img_dir = os.path.join(temp_dir, 'colored_0')
os.makedirs(flow_occ_dir, exist_ok=True)
os.makedirs(img_dir, exist_ok=True)

# Create dummy PNG files for flow and images
# We'll create simple valid PNG file stubs
def create_dummy_png(filepath):
    # Minimal valid PNG file (1x1 white pixel)
    png_data = bytes([
        0x89, 0x50, 0x4E, 0x47, 0x0D, 0x0A, 0x1A, 0x0A,  # PNG signature
        0x00, 0x00, 0x00, 0x0D,  # IHDR chunk size
        0x49, 0x48, 0x44, 0x52,  # IHDR
        0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01,
        0x08, 0x02, 0x00, 0x00, 0x00, 0x90, 0x77, 0x53,
        0xDE,
        0x00, 0x00, 0x00, 0x0C,  # IDAT chunk size
        0x49, 0x44, 0x41, 0x54,  # IDAT
        0x08, 0xD7, 0x63, 0xF8, 0xFF, 0xFF, 0xFF, 0x7F,
        0x00, 0x09, 0xFB, 0x03, 0xFD,
        0x00, 0x00, 0x00, 0x00,  # IEND chunk size
        0x49, 0x45, 0x4E, 0x44,  # IEND
        0xAE, 0x42, 0x60, 0x82
    ])
    with open(filepath, 'wb') as f:
        f.write(png_data)

# Create flow map files
flow_files = ['000000_10.png', '000001_10.png']
for flow_file in flow_files:
    create_dummy_png(os.path.join(flow_occ_dir, flow_file))

# Create corresponding image files
for flow_file in flow_files:
    base_name = flow_file.replace('_10.png', '')
    create_dummy_png(os.path.join(img_dir, base_name + '_10.png'))
    create_dummy_png(os.path.join(img_dir, base_name + '_11.png'))

# Define split2list function that the make_dataset calls
def split2list(images, split, default_split=0.9):
    """Dummy implementation to split the dataset"""
    if split == 'train':
        split_ratio = default_split
    elif split == 'val':
        split_ratio = 1 - default_split
    else:
        split_ratio = default_split
    
    split_point = int(len(images) * split_ratio)
    if split == 'train':
        return images[:split_point]
    else:
        return images[split_point:]

# Inject the function into the module scope
import sys
current_module = sys.modules[__name__]
setattr(current_module, 'split2list', split2list)

ARGS = {
    'dir': temp_dir,
    'split': 'train',
    'occ': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list output has no array shape of its own; none is 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
make_dataset Ran
ran on a synthesized input
McDo/LightFlowPytorch/datasets/mpisintel.py
code served (permissive licence) · get_code("bd1431a30359f7b7")
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. 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 os
import tempfile
import shutil

# Create a temporary directory structure for the dataset
temp_dir = tempfile.mkdtemp()
dataset_dir = temp_dir

# Create flow directory with nested structure
flow_base = os.path.join(dataset_dir, 'flow')
os.makedirs(flow_base)
scene_dir_path = os.path.join(flow_base, 'scene_0')
os.makedirs(scene_dir_path)

# Create a dummy .flo file
flo_file = os.path.join(scene_dir_path, 'flow_0000.flo')
with open(flo_file, 'wb') as f:
    f.write(b'\x00' * 100)

# Create clean image directory with corresponding images
img_base = os.path.join(dataset_dir, 'clean')
os.makedirs(img_base)
scene_img_dir = os.path.join(img_base, 'scene_0')
os.makedirs(scene_img_dir)

# Create dummy image files
for frame_num in [0, 1]:
    img_file = os.path.join(scene_img_dir, 'flow_{:04d}.png'.format(frame_num))
    with open(img_file, 'wb') as f:
        f.write(b'\x89PNG\r\n\x1a\n' + b'\x00' * 50)

# Mock the split2list function since it's called but not defined
def split2list(images, split, default_split=0.87):
    return images

# Inject the mock function into the module namespace
import sys
current_module = sys.modules[__name__]
current_module.split2list = split2list

ARGS = {
    'dataset_dir': dataset_dir,
    'split': 'train',
    'dataset_type': 'clean'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list output has no array shape of its own; none is 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
read_flow Ran
ran on a synthesized input
McDo/LightFlowPytorch/utils/flowlib.py
code served (permissive licence) · get_code("9c84ff434efaba0b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (3, 4, 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 tempfile
import os

# Create a temporary .flo file with valid optical flow data
temp_dir = tempfile.mkdtemp()
filename = os.path.join(temp_dir, "test_flow.flo")

# Write a valid .flo file
with open(filename, 'wb') as f:
    # Write magic number
    magic = np.array([202021.25], dtype=np.float32)
    magic.tofile(f)
    
    # Write height and width
    h = np.array([4], dtype=np.int32)
    w = np.array([3], dtype=np.int32)
    h.tofile(f)
    w.tofile(f)
    
    # Write optical flow data (2 channels for each pixel)
    data = np.random.randn(2 * 3 * 4).astype(np.float32)
    data.tofile(f)

ARGS = {
    "filename": filename
}
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 (3, 4, 2) — shape (3, 4, 2) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
split2list Ran
ran on a synthesized input
McDo/LightFlowPytorch/datasets/util.py
code served (permissive licence) · get_code("1fe407a51fa0b922")
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. 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

np.random.seed(0)

# Create a simple list of image samples (can be anything, the function just iterates)
images = ['image_1.jpg', 'image_2.jpg', 'image_3.jpg', 'image_4.jpg']

# Test with split as a float value
split = 0.75

ARGS = {
    'images': images,
    'split': split,
    'default_split': 0.9
}
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 tuple — a tuple output has no array shape of its own; none is 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
str2bool Ran
ran, violated the contract we drafted
McDo/LightFlowPytorch/utils/util.py
code served (permissive licence) · get_code("25c7475539e39da4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned bool, 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

def str2bool(v):
    return v.lower() in ('yes', 'true', 't', '1')

ARGS = {
    'v': 'true'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, 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
KITTI_loader Not yet run
unverified (harvested, no recorded run)
McDo/LightFlowPytorch/datasets/KITTI.py
code served (permissive licence) · get_code("4e7beea709169af7")
How it ran

no invocation recorded.

flownetc Not yet run
unverified (harvested, no recorded run)
ClementPinard/FlowNetPytorch/models/FlowNetC.py
code served (permissive licence) · get_code("2e9326c5343dd6d0")
How it ran

no invocation recorded.

flownetc_bn Not yet run
unverified (harvested, no recorded run)
ClementPinard/FlowNetPytorch/models/FlowNetC.py
code served (permissive licence) · get_code("59b45a46f776690d")
How it ran

no invocation recorded.

flownetl Not yet run
unverified (harvested, no recorded run)
McDo/LightFlowPytorch/models/FlowNetL.py
code served (permissive licence) · get_code("83d924e4f07e21c6")
How it ran

no invocation recorded.

flying_chairs Not yet run
unverified (harvested, no recorded run)
McDo/LightFlowPytorch/datasets/flyingchairs.py
code served (permissive licence) · get_code("e851f9b7068b4e4a")
How it ran

no invocation recorded.

load_flow_from_png Not yet run
unverified (harvested, no recorded run)
McDo/LightFlowPytorch/datasets/KITTI.py
code served (permissive licence) · get_code("8f05b2af27a4b333")
How it ran

no invocation recorded.

make_dataset Not yet run
unverified (harvested, no recorded run)
McDo/LightFlowPytorch/datasets/flyingchairs.py
code served (permissive licence) · get_code("5b483f6b2c434ae9")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 20 distinct functions with an identity, 2 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 2 later) and 18 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
str2bool 23 other papers
2 dated earlier, 21 later
1410.5401 (2014-10)
Neural Turing Machines
1502.04623 (2015-02)
DRAW: A Recurrent Neural Network For Image Generation
1608.00530 (2016-08)
Early Methods for Detecting Adversarial Images
and 20 more not named here
load_flow_from_png 1 other paper
1 later
2104.00556 (2021-04)
Deep Two-View Structure-from-Motion Revisited

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

Convolutional neural networks (CNNs) have recently been very successful in a variety of computer vision tasks, especially on those linked to recognition. Optical flow estimation has not been among the tasks where CNNs were successful. In this paper we construct appropriate CNNs which are capable of solving the optical flow estimation problem as a supervised learning task. We propose and compare two architectures: a generic architecture and another one including a layer that correlates feature vectors at different image locations. Since existing ground truth data sets are not sufficiently large to train a CNN, we generate a synthetic Flying Chairs dataset. We show that networks trained on this unrealistic data still generalize very well to existing datasets such as Sintel and KITTI, achieving competitive accuracy at frame rates of 5 to 10 fps.

For agents

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

get_harvested_code_for_paper("1504.06852")
get_code_for_paper("1504.06852")
have("1504.06852")

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

curl https://syntology.ai/api/ran/1504.06852.json

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

[![Syntology run record](https://syntology.ai/api/ran/1504.06852.svg)](https://syntology.ai/paper/1504.06852)

Connect an agent — have() is free.