SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2204.03444 · CVPR · 2022

Deep Visual Geo-localization Benchmark

Gabriela Csurka, Riccardo Mereu, Barbara Caputo, Gabriele Trivigno, Gabriele Berton, Carlo Masone, Torsten Sattler

arXiv · PDF · Open in the Atlas

Code that ran

7 of 13 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 6 failures of our instrument; 6 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/2204.03444.json · badge: /api/ran/2204.03444.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
gmberton/deep-visual-geo-localization-benchmark canonical (classifier label)
classifier basis not recorded on the link
7 ran · 6 no recorded run
FunctionStatusWhere it lives
collate_fn Ran
ran; our contract draft was wrong, not the code
gmberton/deep-visual-geo-localization-benchmark/datasets_ws.py
code served (permissive licence) · get_code("c160834241fc1797")
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 torch

torch.manual_seed(0)

# Create a single batch element: (images, triplets_local_indexes, triplets_global_indexes)
# images: (12, 3, h, w) - 12 images with 3 channels, height 4, width 5
images_1 = torch.randn(12, 3, 4, 5)
# triplets_local_indexes: (10, 3) - 10 triplets with 3 indexes each
triplets_local_indexes_1 = torch.randint(0, 12, (10, 3))
# triplets_global_indexes: (12,) - 12 global indexes
triplets_global_indexes_1 = torch.arange(12)

batch_element_1 = (images_1, triplets_local_indexes_1, triplets_global_indexes_1)

# Create a second batch element with same structure
images_2 = torch.randn(12, 3, 4, 5)
triplets_local_indexes_2 = torch.randint(0, 12, (10, 3))
triplets_global_indexes_2 = torch.arange(12, 24)

batch_element_2 = (images_2, triplets_local_indexes_2, triplets_global_indexes_2)

# Create batch as list of tuples
batch = [batch_element_1, batch_element_2]

ARGS = {
    "batch": batch
}
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 — 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
drop_path Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
gmberton/deep-visual-geo-localization-benchmark/model/cct/stochastic_depth.py
code served (permissive licence) · get_code("6bb7492655ecbda4")
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 Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a small tensor with batch size 2 and feature dimension 4
x = torch.randn(2, 4, dtype=torch.float32)

# Define arguments for drop_path function
ARGS = {
    "x": x,
    "drop_prob": 0.5,
    "training": 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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
mac Ran
ran; our contract draft was wrong, not the code
gmberton/deep-visual-geo-localization-benchmark/model/functional.py
code served (permissive licence) · get_code("0b387e5979c78d55")
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, 4, 1, 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 torch
import torch.nn.functional as F

torch.manual_seed(0)

# mac() takes a single parameter x, which is passed to F.adaptive_max_pool2d
# F.adaptive_max_pool2d expects a 4D tensor (batch, channels, height, width)
# Create a small tensor with distinct dimensions: batch=2, channels=4, height=3, width=5
x = torch.randn(2, 4, 3, 5)

ARGS = {
    "x": x
}
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 4, dtype family float, shape (2, 4, 1, 1) — shape (2, 4, 1, 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
path_to_pil_img Ran
ran, honoured the contract we drafted
gmberton/deep-visual-geo-localization-benchmark/datasets_ws.py
code served (permissive licence) · get_code("f0db6d80ebf12bfa")
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 Image. 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 PIL import Image
import io

# Create a minimal valid image as bytes in memory
# Create a small 4x3 RGB image
img = Image.new('RGB', (4, 3), color=(255, 0, 0))
img_bytes = io.BytesIO()
img.save(img_bytes, format='PNG')
img_bytes.seek(0)

# Since the function uses Image.open(path), we need to provide a path
# We'll create a temporary file-like object, but we need an actual path string
# For testing purposes, we'll use a data URI approach or mock

# Actually, let's create a simple approach: use a string path
# Since we can't do actual file I/O, we'll need to mock or use a workaround
# The simplest is to create a temporary file

import tempfile
import os

# Create a temporary directory and file
temp_dir = tempfile.gettempdir()
temp_path = os.path.join(temp_dir, 'test_image.png')

# Create and save a test image to a real temporary file
test_img = Image.new('RGB', (4, 3), color=(255, 0, 0))
test_img.save(temp_path)

ARGS = {
    'path': temp_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
pe_check Ran
ran; our contract draft was wrong, not the code
gmberton/deep-visual-geo-localization-benchmark/model/cct/helpers.py
pointer only (licence: MIT); the stored copy get_code serves, from Lu-Feng/AANet, is pointer only (licence: NONE) · get_code("f9ba92375293c7b1")
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 Lu-Feng/AANet, 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.

sare_ind Ran
ran; our contract draft was wrong, not the code
gmberton/deep-visual-geo-localization-benchmark/model/functional.py
code served (permissive licence) · get_code("463b2f96158e03a8")
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)

# Based on the function signature and body:
# - query: tensor of shape 1xn_features, used in subtraction and sum
# - positive: tensor of shape 1xn_features, used in subtraction
# - negative: tensor of shape 1xn_features, used in subtraction
# The function performs: ((query - positive) ** 2).sum(1) and ((query - negative) ** 2).sum(1)
# Then concatenates results and applies log_softmax

# Create small tensors with shape 1xn_features where n_features=4
n_features = 4

query = torch.randn(1, n_features)
positive = torch.randn(1, n_features)
negative = torch.randn(1, n_features)

ARGS = {
    "query": query,
    "positive": positive,
    "negative": negative
}
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
sare_joint Ran
ran; our contract draft was wrong, not the code
gmberton/deep-visual-geo-localization-benchmark/model/functional.py
code served (permissive licence) · get_code("d15fdc91eec37e29")
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 with distinct sizes
# query: 1 x n_features
query = torch.randn(1, 4)

# positive: 1 x n_features (same n_features as query)
positive = torch.randn(1, 4)

# negatives: n_negative x n_features (multiple negatives, same n_features)
negatives = torch.randn(3, 4)

ARGS = {
    "query": query,
    "positive": positive,
    "negatives": negatives
}
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
cct_2 Not yet run
unverified (harvested, no recorded run)
gmberton/deep-visual-geo-localization-benchmark/model/cct/cct.py
code served (permissive licence) · get_code("cf97fc0d1f953674")
How it ran

no invocation recorded.

cct_4 Not yet run
unverified (harvested, no recorded run)
gmberton/deep-visual-geo-localization-benchmark/model/cct/cct.py
code served (permissive licence) · get_code("f709bdf67ed60e8a")
How it ran

no invocation recorded.

cct_6 Not yet run
unverified (harvested, no recorded run)
gmberton/deep-visual-geo-localization-benchmark/model/cct/cct.py
code served (permissive licence) · get_code("f1b8c16700889467")
How it ran

no invocation recorded.

resize_pos_embed Not yet run
unverified (harvested, no recorded run)
gmberton/deep-visual-geo-localization-benchmark/model/cct/helpers.py
pointer only (licence: MIT); the stored copy get_code serves, from Lu-Feng/AANet, is pointer only (licence: NONE) · get_code("39ae5d524f68a36e")
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 Lu-Feng/AANet, 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. No invocation recorded.

resume_model Not yet run
unverified (harvested, no recorded run)
gmberton/deep-visual-geo-localization-benchmark/util.py
code served (permissive licence) · get_code("69d14231085af9e1")
How it ran

no invocation recorded.

resume_train Not yet run
unverified (harvested, no recorded run)
gmberton/deep-visual-geo-localization-benchmark/util.py
code served (permissive licence) · get_code("8e0b90019577ffc0")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 13 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 1 the same month, 13 later) and 0 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
drop_path 224 other papers
34 dated earlier, 8 the same month, 179 later, 2 same year, order not determinable, 1 undated · 38 only as an assembly brick
2003.12039 (2020-03)
RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
2007.09384 (2020-07)
Multi-Scale Positive Sample Refinement for Few-Shot Object Detection
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
and 221 more not named here
path_to_pil_img 9 other papers
9 later
2306.02994 (2023-06)
Long-range UAV Thermal Geo-localization with Satellite Imagery
2310.05184 (2023-10)
AANet: Aggregation and Alignment Network with Semi-hard Positive Sample Mining for Hierarchical Place Recognition
2402.16086 (2024-02)
Deep Homography Estimation for Visual Place Recognition
and 6 more not named here
collate_fn 7 other papers
7 later
2306.02994 (2023-06)
Long-range UAV Thermal Geo-localization with Satellite Imagery
2310.05184 (2023-10)
AANet: Aggregation and Alignment Network with Semi-hard Positive Sample Mining for Hierarchical Place Recognition
2402.16086 (2024-02)
Deep Homography Estimation for Visual Place Recognition
and 4 more not named here
mac 3 other papers
3 later
2306.02994 (2023-06)
Long-range UAV Thermal Geo-localization with Satellite Imagery
2405.20470 (2024-05)
STHN: Deep Homography Estimation for UAV Thermal Geo-localization with Satellite Imagery
2502.01035 (2025-02)
UASTHN: Uncertainty-Aware Deep Homography Estimation for UAV Satellite-Thermal Geo-localization
resume_model 3 other papers
3 later
2402.19231 (2024-02)
CricaVPR: Cross-image Correlation-aware Representation Learning for Visual Place Recognition
2412.00784 (2024-12)
EDTformer: An Efficient Decoder Transformer for Visual Place Recognition
2509.25723 (2025-09)
SAGE: Spatial-visual Adaptive Graph Exploration for Efficient Visual Place Recognition
resume_train 3 other papers
3 later
2402.19231 (2024-02)
CricaVPR: Cross-image Correlation-aware Representation Learning for Visual Place Recognition
2502.16601 (2025-02)
SelaVPR++: Towards Seamless Adaptation of Foundation Models for Efficient Place Recognition
2509.25723 (2025-09)
SAGE: Spatial-visual Adaptive Graph Exploration for Efficient Visual Place Recognition
sare_ind 3 other papers
3 later
2306.02994 (2023-06)
Long-range UAV Thermal Geo-localization with Satellite Imagery
2405.20470 (2024-05)
STHN: Deep Homography Estimation for UAV Thermal Geo-localization with Satellite Imagery
2502.01035 (2025-02)
UASTHN: Uncertainty-Aware Deep Homography Estimation for UAV Satellite-Thermal Geo-localization
sare_joint 3 other papers
3 later
2306.02994 (2023-06)
Long-range UAV Thermal Geo-localization with Satellite Imagery
2405.20470 (2024-05)
STHN: Deep Homography Estimation for UAV Thermal Geo-localization with Satellite Imagery
2502.01035 (2025-02)
UASTHN: Uncertainty-Aware Deep Homography Estimation for UAV Satellite-Thermal Geo-localization
cct_2 2 other papers
2 later
2310.05184 (2023-10)
AANet: Aggregation and Alignment Network with Semi-hard Positive Sample Mining for Hierarchical Place Recognition
2402.16086 (2024-02)
Deep Homography Estimation for Visual Place Recognition
cct_4 2 other papers
2 later
2310.05184 (2023-10)
AANet: Aggregation and Alignment Network with Semi-hard Positive Sample Mining for Hierarchical Place Recognition
2402.16086 (2024-02)
Deep Homography Estimation for Visual Place Recognition
cct_6 2 other papers
2 later
2310.05184 (2023-10)
AANet: Aggregation and Alignment Network with Semi-hard Positive Sample Mining for Hierarchical Place Recognition
2402.16086 (2024-02)
Deep Homography Estimation for Visual Place Recognition
pe_check 1 other paper
1 later
2310.05184 (2023-10)
AANet: Aggregation and Alignment Network with Semi-hard Positive Sample Mining for Hierarchical Place Recognition
resize_pos_embed 1 other paper
1 later
2310.05184 (2023-10)
AANet: Aggregation and Alignment Network with Semi-hard Positive Sample Mining for Hierarchical Place Recognition

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

In this paper, we propose a new open-source benchmarking framework for Visual Geo-localization (VG) that allows to build, train, and test a wide range of commonly used architectures, with the flexibility to change individual components of a geo-localization pipeline. The purpose of this framework is twofold: i) gaining insights into how different components and design choices in a VG pipeline impact the final results, both in terms of performance (recall@N metric) and system requirements (such as execution time and memory consumption); ii) establish a systematic evaluation protocol for comparing different methods. Using the proposed framework, we perform a large suite of experiments which provide criteria for choosing backbone, aggregation and negative mining depending on the use-case and requirements. We also assess the impact of engineering techniques like pre/post-processing, data augmentation and image resizing, showing that better performance can be obtained through somewhat simple procedures: for example, downscaling the images' resolution to 80% can lead to similar results with a 36% savings in extraction time and dataset storage requirement. Code and trained models are available at dataset storage requirement. https://deep-vg-bench.herokuapp.com/.

For agents

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

get_harvested_code_for_paper("2204.03444")
get_code_for_paper("2204.03444")
have("2204.03444")

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

curl https://syntology.ai/api/ran/2204.03444.json

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

[![Syntology run record](https://syntology.ai/api/ran/2204.03444.svg)](https://syntology.ai/paper/2204.03444)

Connect an agent — have() is free.