SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2006.14613 · NeurIPS · 2020

Space-Time Correspondence as a Contrastive Random Walk

U Berkeley, Allan Jabri, Andrew Owens, Alexei Efros

arXiv · PDF · Open in the Atlas

Code that ran

11 of 14 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 5 have no contract verdict; 6 failures of our instrument; 3 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/2006.14613.json · badge: /api/ran/2006.14613.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
ajabri/videowalk canonical (classifier label)
classifier basis: census
11 ran · 3 no recorded run
FunctionStatusWhere it lives
common_args Ran
ran on a synthesized input
ajabri/videowalk/code/utils/arguments.py
code served (permissive licence) · get_code("cd7e0dea3e1a7fd0")
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 ArgumentParser. 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 argparse

torch_imported = False
try:
    import torch
    torch.manual_seed(0)
    torch_imported = True
except ImportError:
    pass

# Create a parser object
parser = argparse.ArgumentParser(description="Test parser")

ARGS = {
    "parser": parser
}
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 ArgumentParser — a ArgumentParser output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ArgumentParser 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
computePCK Ran
ran on a synthesized input
ajabri/videowalk/code/eval/eval_pck.py
code served (permissive licence) · get_code("11ed14b5cc00d18c")
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, 1) 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

# Build distAll - a dict where keys are indices and values are arrays of distances
# The function iterates through range(len(distAll)), so we need a dict with 
# integer keys 0, 1, 2, etc. and array values containing distance measurements
distAll = {
    0: np.array([0.5, 1.2, 0.3, 2.1, 0.8]),
    1: np.array([1.5, 0.2, 1.8, 0.6, 2.0]),
    2: np.array([0.1, 0.9, 1.1, 0.4, 1.9]),
}

# Set distance threshold for PCK computation
distThresh = 1.0

ARGS = {
    "distAll": distAll,
    "distThresh": distThresh,
}
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, 1) — shape (4, 1) 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_hist Ran
ran on a synthesized input
ajabri/videowalk/code/eval/eval_vip.py
code served (permissive licence) · get_code("3acc87c8ff9403a5")
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 (20, 20) 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 PIL import Image
import tempfile
import os

# Create temporary directory for test files
tmpdir = tempfile.mkdtemp()

# Create a simple test image (4x3 pixels, values 0-19 for class labels)
img_array = np.random.randint(0, 20, size=(4, 3), dtype=np.uint8)
img_path_1 = os.path.join(tmpdir, 'test_img_1.jpg')
img = Image.fromarray(img_array)
img.save(img_path_1)

# Create corresponding label (4x3 pixels with class labels 0-19)
label_array = np.random.randint(0, 20, size=(4, 3), dtype=np.uint8)
label_path_1 = os.path.join(tmpdir, 'test_img_1.png')
label_img = Image.fromarray(label_array)
label_img.save(label_path_1)

# Create second test image and label
img_array_2 = np.random.randint(0, 20, size=(4, 3), dtype=np.uint8)
img_path_2 = os.path.join(tmpdir, 'test_img_2.jpg')
img_2 = Image.fromarray(img_array_2)
img_2.save(img_path_2)

label_array_2 = np.random.randint(0, 20, size=(4, 3), dtype=np.uint8)
label_path_2 = os.path.join(tmpdir, 'test_img_2.png')
label_img_2 = Image.fromarray(label_array_2)
label_img_2.save(label_path_2)

# Define arguments for compute_hist function
ARGS = {
    'images': [img_path_1, img_path_2],
    'labels': [label_path_1, label_path_2]
}
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 ndarray, ndim 2, dtype family float, shape (20, 20) — shape (20, 20) 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
fast_hist Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ajabri/videowalk/code/eval/eval_vip.py
code served (permissive licence) · get_code("d86fb168246b16a4")
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 ndarray, ndim 2, dtype family int, 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

# Analyze the function:
# - a: array-like, used with >= 0, < n comparisons, and .astype(int)
# - b: array-like, same shape as a, used in indexing
# - n: int, used as upper bound and for bincount minlength

# Create small numpy arrays for a and b
a = np.array([0, 1, 2, 3, 1, 2], dtype=np.int32)
b = np.array([0, 1, 2, 0, 1, 2], dtype=np.int32)
n = 4

ARGS = {
    "a": a,
    "b": b,
    "n": n
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 int, 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
get_frame_aug Ran
ran; our contract draft was wrong, not the code
ajabri/videowalk/code/utils/augs.py
pointer only (licence: MIT); the stored copy get_code serves, from shvdiwnkozbw/ssl-uvos, is pointer only (licence: NONE) · get_code("fb2519bb639a667c")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from shvdiwnkozbw/ssl-uvos, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

make_gif Ran
ran on a synthesized input
ajabri/videowalk/code/utils/visualize.py
code served (permissive licence) · get_code("68336aa7c98a4d33")
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 4, dtype family int, shape (2, 256, 256, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Create a small test video tensor
# Shape: (T, C, H, W) where T=2 frames, C=3 channels, H=4, W=5
video_tensor = torch.rand(2, 3, 4, 5)

ARGS = {
    'video': video_tensor,
    'outname': None,
    'sz': 256
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, imageio, 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 4, dtype family int, shape (2, 256, 256, 3) — shape (2, 256, 256, 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
n_patches Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ajabri/videowalk/code/utils/augs.py
pointer only (licence: MIT); the stored copy get_code serves, from shvdiwnkozbw/ssl-uvos, is pointer only (licence: NONE) · get_code("14b1cc74dcb694bf")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from shvdiwnkozbw/ssl-uvos, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

patch_grid Ran
ran; our contract draft was wrong, not the code
ajabri/videowalk/code/utils/augs.py
pointer only (licence: MIT); the stored copy get_code serves, from shvdiwnkozbw/ssl-uvos, is pointer only (licence: NONE) · get_code("600497c0fb4eb8c7")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from shvdiwnkozbw/ssl-uvos, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

pca_feats Ran
ran on a synthesized input
ajabri/videowalk/code/utils/visualize.py
code served (permissive licence) · get_code("1b4515fda7085b63")
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, 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 numpy as np
from sklearn.decomposition import PCA

torch.manual_seed(0)

# Create a small feature tensor with shape (N, C, H, W)
# N=2 (batch), C=4 (channels), H=3 (height), W=5 (width)
ff = torch.randn(2, 4, 3, 5)

ARGS = {
    'ff': ff,
    'K': 1,
    'solver': 'auto',
    'whiten': True,
    'img_normalize': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, 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, 3, 5) — shape (2, 3, 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
resnet18 Ran
ran; our contract draft was wrong, not the code
ajabri/videowalk/code/resnet.py
code served (permissive licence) · get_code("4a4432bf960d0dbe")
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 ResNet. 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
from torchvision.models.resnet import BasicBlock

torch.manual_seed(0)

# Define mock load_state_dict_from_url since it's not imported
def load_state_dict_from_url(url, progress=True):
    # Return empty dict as we won't actually load anything
    return {}

# Patch the function into the module namespace for the function to work
import sys
current_module = sys.modules[__name__]
current_module.load_state_dict_from_url = load_state_dict_from_url

ARGS = {
    'pretrained': False,
    'progress': True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
resnet50 Ran
ran; our contract draft was wrong, not the code
ajabri/videowalk/code/resnet.py
code served (permissive licence) · get_code("94b05476d48954e9")
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 ResNet. 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
from torchvision.models.resnet import Bottleneck

torch.manual_seed(0)

ARGS = {
    'pretrained': False,
    'progress': True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
draw_matches Not yet run
unverified (harvested, no recorded run)
ajabri/videowalk/code/utils/visualize.py
code served (permissive licence) · get_code("fe5e025dcc4db6f3")
How it ran

no invocation recorded.

get_frame_transform Not yet run
unverified (harvested, no recorded run)
ajabri/videowalk/code/utils/kornia_augs.py
code served (permissive licence) · get_code("57e8ff198fc0ac02")
How it ran

no invocation recorded.

patch_grid Not yet run
unverified (harvested, no recorded run)
ajabri/videowalk/code/utils/kornia_augs.py
code served (permissive licence) · get_code("b4eaa298ffce395e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 6 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 1 the same month, 6 later) and 8 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
resnet18 95 other papers
11 dated earlier, 1 the same month, 82 later, 1 undated
1805.00907 (2018-05)
Glow: Graph Lowering Compiler Techniques for Neural Networks
1902.08153 (2019-02)
Learned Step Size Quantization
1904.01355 (2019-04)
FCOS: Fully Convolutional One-Stage Object Detection
and 92 more not named here
fast_hist 12 other papers
9 dated earlier, 3 later
1511.00561 (2015-11)
SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation
1702.01105 (2017-02)
Joint 2D-3D-Semantic Data for Indoor Scene Understanding
1806.07373 (2018-06)
Few-Shot Segmentation Propagation with Guided Networks
and 9 more not named here
get_frame_aug 2 other papers
2 later
2308.09951 (2023-08)
Semantics Meets Temporal Correspondence: Self-supervised Object-centric Learning in Videos
2311.17893 (2023-11)
Betrayed by Attention: A Simple yet Effective Approach for Self-supervised Video Object Segmentation
n_patches 2 other papers
2 later
2308.09951 (2023-08)
Semantics Meets Temporal Correspondence: Self-supervised Object-centric Learning in Videos
2311.17893 (2023-11)
Betrayed by Attention: A Simple yet Effective Approach for Self-supervised Video Object Segmentation
patch_grid 2 other papers
2 later
2308.09951 (2023-08)
Semantics Meets Temporal Correspondence: Self-supervised Object-centric Learning in Videos
2311.17893 (2023-11)
Betrayed by Attention: A Simple yet Effective Approach for Self-supervised Video Object Segmentation
resnet50 1 other paper
1 later
2107.02156 (2021-07)
Do Different Tracking Tasks Require Different Appearance Models?

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

Figure 1: We represent video as a graph, where nodes are image patches, and edges are affinities (in some feature space) between nodes of neighboring frames. Our aim is to learn features such that temporal correspondences are represented by strong edges. We learn to find paths through the graph by performing a random walk between query and target nodes. A contrastive loss encourages paths that reach the target, implicitly supervising latent correspondence along the path. Learning proceeds without labels by training on a palindrome sequence, walking from frame t to t + k, then back to t, using the initial node itself as the target. Please see our webpage for videos.

For agents

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

get_harvested_code_for_paper("2006.14613")
get_code_for_paper("2006.14613")
have("2006.14613")

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

curl https://syntology.ai/api/ran/2006.14613.json

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

[![Syntology run record](https://syntology.ai/api/ran/2006.14613.svg)](https://syntology.ai/paper/2006.14613)

Connect an agent — have() is free.