SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2304.02008 · ICCV · 2023

GlueStick: Robust Image Matching by Sticking Points and Lines Together

Marc Pollefeys, Yifan Yu, Rémi Pautrat, Viktor Larsson, Iago Suárez

arXiv · PDF · Open in the Atlas

Code that ran

20 of 29 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (10 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 2 violated it and 11 have no contract verdict; 7 failures of our instrument; 9 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2304.02008.json · badge: /api/ran/2304.02008.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
Golbstein/GlueStick-tensorflow community
Papers with Code archive
11 ran · 4 no recorded run
cvg/GlueStick community
Papers with Code archive
9 ran · 5 no recorded run
FunctionStatusWhere it lives
AttentionalPropagation Ran
ran on a synthesized input · constructed an object; no result was computed
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("8879256fa01366c9")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

AttentionalPropagation Ran
ran on a synthesized input · constructed an object; no result was computed
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("0bf652d5cd14f0d1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, AttentionalPropagation: the run constructed the object and did not call it, so no result was computed. 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 torch import nn
from copy import deepcopy

torch.manual_seed(0)

# Build the function dependencies first
def MLP(channels, do_bn=True):
    n = len(channels)
    layers = []
    for i in range(1, n):
        layers.append(nn.Conv1d(channels[i - 1], channels[i], kernel_size=1, bias=True))
        if i < n - 1:
            if do_bn:
                layers.append(nn.BatchNorm1d(channels[i]))
            layers.append(nn.ReLU())
    return nn.Sequential(*layers)

class MultiHeadedAttention(nn.Module):
    def __init__(self, h, d_model):
        super().__init__()
        assert d_model % h == 0
        self.dim = d_model // h
        self.h = h
        self.merge = nn.Conv1d(d_model, d_model, kernel_size=1)
        self.proj = nn.ModuleList([deepcopy(self.merge) for _ in range(3)])

    def forward(self, query, key, value):
        b = query.size(0)
        query, key, value = [l(x).view(b, self.dim, self.h, -1) for l, x in zip(self.proj, (query, key, value))]
        x, prob = attention(query, key, value)
        return self.merge(x.contiguous().view(b, self.dim * self.h, -1))

@torch.cuda.amp.custom_fwd(cast_inputs=torch.float32)
def attention(query, key, value):
    dim = query.shape[1]
    scores = torch.einsum('bdhn,bdhm->bhnm', query, key) / dim ** 0.5
    prob = torch.nn.functional.softmax(scores, dim=-1)
    return (torch.einsum('bhnm,bdhm->bdhn', prob, value), prob)

class AttentionalPropagation(nn.Module):
    def __init__(self, num_dim, num_heads, skip_init=False):
        super().__init__()
        self.attn = MultiHeadedAttention(num_heads, num_dim)
        self.mlp = MLP([num_dim * 2, num_dim * 2, num_dim], do_bn=True)
        nn.init.constant_(self.mlp[-1].bias, 0.0)
        if skip_init:
            self.register_parameter('scaling', nn.Parameter(torch.tensor(0.0)))
        else:
            self.scaling = 1.0

    def forward(self, x, source):
        message = self.attn(x, source, source)
        return self.mlp(torch.cat([x, message], dim=1)) * self.scaling

# Now create arguments for AttentionalPropagation
ARGS = {
    'num_dim': 4,
    'num_heads': 2,
    'skip_init': False
}
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 AttentionalPropagation — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a AttentionalPropagation 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
EndPtEncoder Ran
ran on a synthesized input · constructed an object; no result was computed
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("91eb969dfeb55794")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

EndPtEncoder Ran
ran on a synthesized input · constructed an object; no result was computed
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("0744c1b780ad0b62")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, EndPtEncoder: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch import nn

torch.manual_seed(0)

ARGS = {
    "feature_dim": 8,
    "layers": [16, 32, 16]
}
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 EndPtEncoder — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a EndPtEncoder 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
GNNLayer Ran
ran on a synthesized input · constructed an object; no result was computed
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("4094d3ffd7fbc53a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

GNNLayer Ran
ran on a synthesized input · constructed an object; no result was computed
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("c8110ff5f4741d3d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, GNNLayer: the run constructed the object and did not call it, so no result was computed. 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
import torch.nn as nn
from copy import deepcopy

torch.manual_seed(0)

# Define the helper classes and functions needed
class MultiHeadedAttention(nn.Module):
    def __init__(self, h, d_model):
        super().__init__()
        assert d_model % h == 0
        self.dim = d_model // h
        self.h = h
        self.merge = nn.Conv1d(d_model, d_model, kernel_size=1)
        self.proj = nn.ModuleList([deepcopy(self.merge) for _ in range(3)])

    def forward(self, query, key, value):
        b = query.size(0)
        query, key, value = [l(x).view(b, self.dim, self.h, -1) for l, x in zip(self.proj, (query, key, value))]
        x, prob = attention(query, key, value)
        return self.merge(x.contiguous().view(b, self.dim * self.h, -1))

@torch.cuda.amp.custom_fwd(cast_inputs=torch.float32)
def attention(query, key, value):
    dim = query.shape[1]
    scores = torch.einsum('bdhn,bdhm->bhnm', query, key) / dim ** 0.5
    prob = torch.nn.functional.softmax(scores, dim=-1)
    return (torch.einsum('bhnm,bdhm->bdhn', prob, value), prob)

def MLP(channels, do_bn=True):
    n = len(channels)
    layers = []
    for i in range(1, n):
        layers.append(nn.Conv1d(channels[i - 1], channels[i], kernel_size=1, bias=True))
        if i < n - 1:
            if do_bn:
                layers.append(nn.BatchNorm1d(channels[i]))
            layers.append(nn.ReLU())
    return nn.Sequential(*layers)

class AttentionalPropagation(nn.Module):
    def __init__(self, num_dim, num_heads, skip_init=False):
        super().__init__()
        self.attn = MultiHeadedAttention(num_heads, num_dim)
        self.mlp = MLP([num_dim * 2, num_dim * 2, num_dim], do_bn=True)
        nn.init.constant_(self.mlp[-1].bias, 0.0)
        if skip_init:
            self.register_parameter('scaling', nn.Parameter(torch.tensor(0.0)))
        else:
            self.scaling = 1.0

    def forward(self, x, source):
        message = self.attn(x, source, source)
        return self.mlp(torch.cat([x, message], dim=1)) * self.scaling

class GNNLayer(nn.Module):
    def __init__(self, feature_dim, layer_type, skip_init):
        super().__init__()
        assert layer_type in ['cross', 'self']
        self.type = layer_type
        self.update = AttentionalPropagation(feature_dim, 4, skip_init)

    def forward(self, desc0, desc1):
        if self.type == 'cross':
            src0, src1 = (desc1, desc0)
        elif self.type == 'self':
            src0, src1 = (desc0, desc1)
        else:
            raise ValueError('Unknown layer type: ' + self.type)
        delta0, delta1 = (self.update(desc0, src0), self.update(desc1, src1))
        desc0, desc1 = (desc0 + delta0, desc1 + delta1)
        return (desc0, desc1)

# Create valid arguments for GNNLayer
ARGS = {
    'feature_dim': 4,
    'layer_type': 'cross',
    'skip_init': False,
}
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 GNNLayer — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a GNNLayer 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
KeypointEncoder Ran
ran on a synthesized input · constructed an object; no result was computed
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("59b08bff5dc85b8f")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

KeypointEncoder Ran
ran on a synthesized input · constructed an object; no result was computed
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("92f7e8a94de3df6a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, KeypointEncoder: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch import nn

torch.manual_seed(0)

# Construct valid arguments for KeypointEncoder
ARGS = {
    "feature_dim": 8,  # Output feature dimension
    "layers": [16, 32, 16],  # Hidden layer dimensions for MLP
}
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 KeypointEncoder — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a KeypointEncoder 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
LineLayer Ran
ran on a synthesized input · constructed an object; no result was computed
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("76d2114282e14498")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

MLP Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("6ee94794a1b527f5")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

MLP Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("54fd8d988f6f0e09")
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)

# The MLP function takes:
# 1. channels: a list of integers representing channel dimensions for Conv1d layers
# 2. do_bn: a boolean flag for whether to use BatchNorm1d

# From reading the body:
# - channels is used with len() and indexing to build Conv1d layers
# - channels[i-1] and channels[i] are passed as in_channels and out_channels to Conv1d
# - do_bn is used in a boolean if statement to conditionally add BatchNorm1d

# Build a simple channel configuration for a small MLP
channels = [4, 8, 16, 3]  # input dim 4, two hidden layers (8, 16), output dim 3
do_bn = True

ARGS = {
    "channels": channels,
    "do_bn": do_bn
}
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
MultiHeadedAttention Ran
ran on a synthesized input
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("0d5fe463a1a289eb")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

MultiHeadedAttention Ran
ran on a synthesized input · constructed an object; no result was computed
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("35811a32bf5387ee")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, MultiHeadedAttention: the run constructed the object and did not call it, so no result was computed. 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 torch import nn
from copy import deepcopy

torch.manual_seed(0)

# MultiHeadedAttention.__init__ takes h and d_model
# h = number of heads
# d_model = model dimension (must be divisible by h)

h = 2
d_model = 4

ARGS = {
    "h": h,
    "d_model": d_model,
}
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 MultiHeadedAttention — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a MultiHeadedAttention 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
arange_like Ran
ran; our contract draft was wrong, not the code
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("a6e4f5a51c5441d3")
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 1, dtype family float, shape (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

torch.manual_seed(0)

# Create a small tensor
x = torch.randn(2, 4)

# Create a dim parameter
dim = 1

ARGS = {
    "x": x,
    "dim": dim,
}
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 1, dtype family float, shape (4,) — shape (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
attention Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("35ddab8839071886")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

dict_to_dataclass Ran
ran; our contract draft was wrong, not the code
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("46fd24631d359569")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

gather_like_in_torch Ran
ran, violated the contract we drafted
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("4273e23e924da510")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

log_double_softmax Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("198c0e35a834307d")
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 (2, 5, 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

torch.manual_seed(0)

# Analyze the function:
# - scores: shape (b, m, n), used in operations expecting 3D tensor
# - bin_score: scalar or 0D tensor, expanded and concatenated with scores
# The function performs log_softmax operations and combines scores

# Create small tensors with distinct dimensions to probe shapes
b, m, n = 2, 4, 3

scores = torch.randn(b, m, n)
bin_score = torch.tensor(0.5)

ARGS = {
    "scores": scores,
    "bin_score": bin_score,
}
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 (2, 5, 4) — shape (2, 5, 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
normalize_keypoints Ran
ran, violated the contract we drafted
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("6c6c724fa54239ed")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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.

normalize_keypoints Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("7a6d49963474000d")
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 (2, 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 torch

torch.manual_seed(0)

# Create keypoints tensor: shape (batch_size, num_keypoints, 2)
# batch=2, num_keypoints=4, coordinates=2 (x, y)
kpts = torch.randn(2, 4, 2)

# Create shape_or_size as a tuple (h, w)
# Use distinct dimensions: height=3, width=5
shape_or_size = (3, 5)

ARGS = {
    "kpts": kpts,
    "shape_or_size": shape_or_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 Tensor, ndim 3, dtype family float, shape (2, 4, 2) — shape (2, 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
AttentionalGNN Not yet run
unverified (harvested, no recorded run)
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("7c4811d20047f201")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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. No invocation recorded.

AttentionalGNN Not yet run
unverified (harvested, no recorded run)
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("2e1f2c4bc7c2ecd9")
How it ran

no invocation recorded.

BaseModel Not yet run
unverified (harvested, no recorded run)
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("d422ebc642067c6e")
How it ran

no invocation recorded.

GlueStick Not yet run
unverified (harvested, no recorded run)
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("baf352f0d6611c26")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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. No invocation recorded.

GlueStick Not yet run
unverified (harvested, no recorded run)
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("7edc3e49a3a70c3c")
How it ran

no invocation recorded.

LineLayer Not yet run
unverified (harvested, no recorded run)
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("3c20b46867b9cf11")
How it ran

no invocation recorded.

attention Not yet run
unverified (harvested, no recorded run)
cvg/GlueStick/gluestick/models/gluestick.py
code served (permissive licence) · get_code("5317c011246420e0")
How it ran

no invocation recorded.

log_double_softmax Not yet run
unverified (harvested, no recorded run)
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("7e2f151d8d465b17")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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. No invocation recorded.

scatter_mean Not yet run
unverified (harvested, no recorded run)
Golbstein/GlueStick-tensorflow/gluestick.py
pointer only (licence: NONE) · get_code("403889105a73bc9b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Golbstein/GlueStick-tensorflow 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. No invocation recorded.

Also present in other papers' code

Of 29 samples lifted from this paper's linked repositories, 29 carry a function identity and 0 do not. Of the 29 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 28 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
BaseModel
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2103.09213 (2021-03)
Back to the Feature: Learning Robust Camera Localization from Pixels to Pose

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

Line segments are powerful features complementary to points. They offer structural cues, robust to drastic viewpoint and illumination changes, and can be present even in texture-less areas. However, describing and matching them is more challenging compared to points due to partial occlusions, lack of texture, or repetitiveness. This paper introduces a new matching paradigm, where points, lines, and their descriptors are unified into a single wireframe structure. We propose GlueStick, a deep matching Graph Neural Network (GNN) that takes two wireframes from different images and leverages the connectivity information between nodes to better glue them together. In addition to the increased efficiency brought by the joint matching, we also demonstrate a large boost of performance when leveraging the complementary nature of these two features in a single architecture. We show that our matching strategy outperforms the state-of-the-art approaches independently matching line segments and points for a wide variety of datasets and tasks. The code is available at https: //github.com/cvg/GlueStick.

For agents

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

get_harvested_code_for_paper("2304.02008")
get_code_for_paper("2304.02008")
have("2304.02008")

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

curl https://syntology.ai/api/ran/2304.02008.json

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

[![Syntology run record](https://syntology.ai/api/ran/2304.02008.svg)](https://syntology.ai/paper/2304.02008)

Connect an agent — have() is free.