SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.11095 · ICLR · 2024

GIM: Learning Generalizable Image Matcher From Internet Videos

Cheng Wang, Matthias Üller, Wei Yin, Zhipeng Cai, Xuelun Shen, Zijun Li, Kaixuan Wang, Xiaozhi Chen

arXiv · PDF · Open in the Atlas

Code that ran

15 of 17 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; 7 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 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/2402.11095.json · badge: /api/ran/2402.11095.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
xuelunshen/gim canonical (classifier label)
classifier basis: census
15 ran · 2 no recorded run
FunctionStatusWhere it lives
assign_keypoints Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
xuelunshen/gim/hloc/match_dense.py
code served (permissive licence) · get_code("6b4df3b8d1f95653")
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 1, dtype family int, shape (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 numpy as np
from collections import Counter
from typing import List, Tuple, Optional, Union

# Set random seed for reproducibility
np.random.seed(0)

# Create sample keypoints - 5 keypoints with 2D coordinates
kpts = np.array([
    [10.3, 20.5],
    [15.2, 25.1],
    [30.7, 40.2],
    [50.1, 60.3],
    [100.5, 110.2]
], dtype=np.float32)

# Create other_cpts as a list of tuples for non-update case
other_cpts_list = [
    (10.0, 20.0),
    (15.0, 25.0),
    (31.0, 40.0),
]

# Create other_cpts as list of tuples for update case
other_cpts_for_update = [
    (10.0, 20.0),
    (15.0, 25.0),
]

# Set max_error threshold
max_error = 2.0

# Create scores array (same length as kpts)
scores = np.array([0.8, 0.9, 0.7, 0.95, 0.6], dtype=np.float32)

# Create ref_bins for update case (list of Counters)
ref_bins = [Counter(), Counter()]

# Cell size for quantization
cell_size = 5

# Build ARGS dictionary for the function call
ARGS = {
    "kpts": kpts,
    "other_cpts": other_cpts_list,
    "max_error": max_error,
    "update": False,
    "ref_bins": None,
    "scores": None,
    "cell_size": None,
}

# Alternative configuration for update=True case
ARGS_UPDATE = {
    "kpts": kpts,
    "other_cpts": other_cpts_for_update,
    "max_error": max_error,
    "update": True,
    "ref_bins": ref_bins,
    "scores": scores,
    "cell_size": cell_size,
}

# Use the non-update case as the primary ARGS
ARGS = {
    "kpts": kpts,
    "other_cpts": other_cpts_list,
    "max_error": max_error,
    "update": False,
    "ref_bins": None,
    "scores": None,
    "cell_size": None,
}
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 1, dtype family int, shape (5,) — shape (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
build_augmentor Ran
ran; our contract draft was wrong, not the code
xuelunshen/gim/datasets/augment.py
code served (permissive licence) · get_code("a6b2a9c667b63a4c")
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 DarkAug. 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 numpy as np
import albumentations as A

torch.manual_seed(0)

# The build_augmentor function takes:
# - method: a string or None specifying which augmentor to build
# - **kwargs: additional keyword arguments (not used in the function body)

# Test with method='dark' to exercise the DarkAug path
ARGS = {
    'method': 'dark'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
albumentations, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DarkAug — a DarkAug output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DarkAug 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
camera_intrinsic_transform Ran
ran, honoured the contract we drafted
xuelunshen/gim/datasets/scenenet/utils.py
code served (permissive licence) · get_code("afe48ee65c2a83aa")
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, 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 math
import numpy as np

ARGS = {
    "vfov": 45,
    "hfov": 60,
    "pixel_width": 320,
    "pixel_height": 240
}
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 (3, 4) — shape (3, 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
collate_fn Ran
ran; our contract draft was wrong, not the code
xuelunshen/gim/datasets/data.py
pointer only (licence: MIT); the stored copy get_code serves, from ciodar/universalattribution, is pointer only (licence: NONE) · get_code("12c09a1a10ca585d")
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 ciodar/universalattribution, 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.

get_divisible_wh Ran
ran, honoured the contract we drafted
xuelunshen/gim/video_preprocessor.py
code served (permissive licence) · get_code("3e3aed34677c7943")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    "w": 640,
    "h": 480,
    "df": 32
}
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 tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_grouped_ids Ran
ran; our contract draft was wrong, not the code
xuelunshen/gim/hloc/match_dense.py
code served (permissive licence) · get_code("903ef22b452ab670")
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 numpy as np

# get_grouped_ids takes an array parameter
# It uses np.argsort, np.unique with return_counts and return_index, and np.split
# The function groups indices by unique values in the array

# Create a small test array with some repeated values
array = np.array([3, 1, 2, 1, 3, 2, 1])

ARGS = {
    "array": array
}
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 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
get_resized_wh Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
xuelunshen/gim/video_preprocessor.py
code served (permissive licence) · get_code("2bcc62914d8b4ce5")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

ARGS = {
    "w": 640,
    "h": 480,
    "resize": (1080, 1920)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_resized_wh Ran
ran, honoured the contract we drafted
xuelunshen/gim/datasets/utils.py
code served (permissive licence) · get_code("21abf193b3d22564")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - w: width (int or numeric)
# - h: height (int or numeric)
# - resize: optional resize parameter (None or numeric value)
# The function computes a scaling factor and returns resized dimensions

ARGS = {
    "w": 800,
    "h": 600,
    "resize": 400
}
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 tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
imread_color Ran
ran, honoured the contract we drafted
xuelunshen/gim/datasets/utils.py
code served (permissive licence) · get_code("c736348f2c26f6a4")
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 (10, 15, 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 cv2
import numpy as np
import tempfile
import os

# Create a temporary test image file
temp_dir = tempfile.mkdtemp()
test_image_path = os.path.join(temp_dir, "test_image.png")

# Create a simple BGR image (OpenCV uses BGR format)
test_image = np.random.randint(0, 256, (10, 15, 3), dtype=np.uint8)
cv2.imwrite(test_image_path, test_image)

# Define a simple augmentation function that takes an RGB image and returns it
def simple_augment(image):
    """Simple augmentation function that takes RGB image and returns RGB image"""
    return image

ARGS = {
    "path": test_image_path,
    "augment_fn": simple_augment
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
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 3, dtype family int, shape (10, 15, 3) — shape (10, 15, 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_gray Ran
ran, honoured the contract we drafted
xuelunshen/gim/datasets/utils.py
code served (permissive licence) · get_code("28b32b7ce4eceafa")
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 int, shape (32, 32) 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 cv2
import numpy as np
import tempfile
import os

# Create a temporary grayscale image file
temp_dir = tempfile.mkdtemp()
temp_image_path = os.path.join(temp_dir, "test_image.png")

# Create a simple grayscale test image
gray_image = np.random.randint(0, 256, (32, 32), dtype=np.uint8)
cv2.imwrite(temp_image_path, gray_image)

# Define an augmentation function that takes an RGB image and returns an RGB image
def augment_fn(image):
    """Simple augmentation function that takes RGB and returns RGB"""
    # Apply a simple brightness adjustment
    return np.clip(image.astype(np.float32) * 1.1, 0, 255).astype(np.uint8)

# Construct the arguments dictionary
ARGS = {
    "path": temp_image_path,
    "augment_fn": augment_fn
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
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 int, shape (32, 32) — shape (32, 32) 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
position_to_np_array Ran
ran, honoured the contract we drafted
xuelunshen/gim/datasets/scenenet/utils.py
code served (permissive licence) · get_code("ef0704c1e391fa96")
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 1, dtype family float, shape (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

class Position:
    def __init__(self, x, y, z):
        self.x = x
        self.y = y
        self.z = z

position = Position(x=1.0, y=2.5, z=3.7)

ARGS = {
    "position": position,
    "homogenous": 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 ndarray, ndim 1, dtype family float, shape (3,) — shape (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
read_deeplab_image Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
xuelunshen/gim/video_preprocessor.py
code served (permissive licence) · get_code("f983f6b4bdc58b8d")
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 (1, 3, 100, 80) 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 cv2
import torch
import numpy as np

torch.manual_seed(0)

# Create a small test image: 3D numpy array with shape (height, width, channels)
# Using distinct sizes: 100x80x3 so we can probe dimensions later
img = np.random.randint(0, 256, (100, 80, 3), dtype=np.uint8)

ARGS = {
    "img": img,
    "size": 1920
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (1, 3, 100, 80) — shape (1, 3, 100, 80) 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
read_depth Ran
ran, honoured the contract we drafted
xuelunshen/gim/datasets/scenenet/utils.py
code served (permissive licence) · get_code("34e02c0f95a179a4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (2, 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
from unittest.mock import Mock, patch
import tempfile
import os

# Create a temporary image file for testing
temp_dir = tempfile.mkdtemp()
temp_filename = os.path.join(temp_dir, "depth_test.png")

# Create a small depth image (16-bit grayscale) with values in mm
# Using uint16 since depth is typically stored as uint16
depth_data = np.array([[1000, 2000], [3000, 4000]], dtype=np.uint16)

# Save using a simple PNG-like approach via numpy
from PIL import Image
img = Image.fromarray(depth_data, mode='I;16')
img.save(temp_filename)

ARGS = {
    "filename": temp_filename
}
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 2, dtype family float, shape (2, 2) — shape (2, 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
resize_image Ran
ran; our contract draft was wrong, not the code
xuelunshen/gim/hloc/extract_features.py
code served (permissive licence) · get_code("47ffaeadde70c45c")
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 (5, 8, 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 cv2
import numpy as np
import PIL.Image

# Create a small test image (4 rows, 6 columns, 3 channels for RGB)
image = np.random.randint(0, 256, size=(4, 6, 3), dtype=np.uint8)

# Define target size (width, height)
size = (8, 5)

# Test with cv2 interpolation method
interp = 'cv2_linear'

ARGS = {
    'image': image,
    'size': size,
    'interp': interp
}
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 3, dtype family int, shape (5, 8, 3) — shape (5, 8, 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
to_cpts Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
xuelunshen/gim/hloc/match_dense.py
code served (permissive licence) · get_code("e54750836225cbe0")
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, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct valid arguments for to_cpts function
# The function takes:
# - kpts: a 2D array-like of keypoints (gets processed with arithmetic operations)
# - ps: a scalar (float) that represents a quantization parameter

# Create a small array of keypoints (2D coordinates)
kpts = np.array([
    [1.5, 2.3],
    [3.7, 4.2],
    [5.1, 6.9],
    [7.3, 8.5]
])

# ps is a positive float parameter (quantization step)
ps = 0.5

ARGS = {
    "kpts": kpts,
    "ps": ps
}
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 list, ndim 2, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
error_auc Not yet run
unverified (harvested, no recorded run)
xuelunshen/gim/analysis.py
code served (permissive licence) · get_code("7580602299bba28c")
How it ran

no invocation recorded.

main Not yet run
unverified (harvested, no recorded run)
xuelunshen/gim/hloc/extract_features.py
code served (permissive licence) · get_code("5f12182269ed0b51")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 4 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 4 later) and 13 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
collate_fn 15 other papers
8 dated earlier, 7 later
2003.04618 (2020-03)
Convolutional Occupancy Networks
2209.09552 (2022-09)
Cross-modal Learning for Image-Guided Point Cloud Shape Completion
2301.13823 (2023-01)
Grounding Language Models to Images for Multimodal Inputs and Outputs
and 12 more not named here
get_grouped_ids 1 other paper
1 later
2601.04185 (2026-01)
ImLoc: Revisiting Visual Localization with Image-based Representation
resize_image 1 other paper
1 later
2601.04185 (2026-01)
ImLoc: Revisiting Visual Localization with Image-based Representation
to_cpts 1 other paper
1 later
2601.04185 (2026-01)
ImLoc: Revisiting Visual Localization with Image-based Representation

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

Training image matching models requires multi-view images and ground-truth correspondences. Data diversity and scale have been the key towards generalizable models in other computer vision problems (Radford et al., 2021;Ranftl et al., 2020;Kirillov et al., 2023). Inspired by this observation, we propose GIM (Fig. 2), a self-training framework utilizing internet videos to learn a single generalizable model based on any image matching architecture. City ( 43) Scenario (39)

For agents

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

get_harvested_code_for_paper("2402.11095")
get_code_for_paper("2402.11095")
have("2402.11095")

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

curl https://syntology.ai/api/ran/2402.11095.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.11095.svg)](https://syntology.ai/paper/2402.11095)

Connect an agent — have() is free.