SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1907.07484 · 2019

Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Coming

arXiv · PDF · Open in the Atlas

Code that ran

5 of 5 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 3 have no contract verdict; 1 failure of our instrument; 0 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/1907.07484.json · badge: /api/ran/1907.07484.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
bethgelab/robust-detection-benchmark canonical (classifier label)
classifier basis: census
3 ran · 0 no recorded run
bethgelab/imagecorruptions canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
FunctionStatusWhere it lives
clipped_zoom Ran
ran; our contract draft was wrong, not the code
bethgelab/imagecorruptions/imagecorruptions/corruptions.py
code served (permissive licence) · get_code("9f1bb16740f8d1ed")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (8, 6, 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
from scipy.ndimage import zoom as scizoom

# Create a small test image with shape (height, width, channels)
# Using distinct dimensions: 8x6x3 to make shapes unambiguous
img = np.random.rand(8, 6, 3).astype(np.float32)

# zoom_factor is a float that scales the image
zoom_factor = 2.0

ARGS = {
    "img": img,
    "zoom_factor": zoom_factor,
}
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 3, dtype family float, shape (8, 6, 3) — shape (8, 6, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_coco_style_results Ran
ran on a synthesized input
bethgelab/robust-detection-benchmark/data-analysis/robustness_eval.py
code served (permissive licence) · get_code("d76f63e8bd69d6b6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (15, 6, 12) 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 pickle
import numpy as np
import tempfile
import os

# Create a temporary pickle file with mock evaluation output
# The structure should match what get_coco_style_results expects
eval_output = {}

# Create mock data for 15 distortions (benchmark uses first 15)
for distortion_idx in range(15):
    distortion_name = f'distortion_{distortion_idx}'
    eval_output[distortion_name] = {}
    
    # Severity levels 0-5 (0 is clean, 1-5 are corrupted)
    for severity in range(6):
        eval_output[distortion_name][severity] = {
            'bbox': {
                'AP': 0.5 + np.random.rand() * 0.1,
                'AP50': 0.6 + np.random.rand() * 0.1,
                'AP75': 0.55 + np.random.rand() * 0.1,
                'APs': 0.3 + np.random.rand() * 0.1,
                'APm': 0.5 + np.random.rand() * 0.1,
                'APl': 0.65 + np.random.rand() * 0.1,
                'AR1': 0.4 + np.random.rand() * 0.1,
                'AR10': 0.5 + np.random.rand() * 0.1,
                'AR100': 0.55 + np.random.rand() * 0.1,
                'ARs': 0.35 + np.random.rand() * 0.1,
                'ARm': 0.52 + np.random.rand() * 0.1,
                'ARl': 0.68 + np.random.rand() * 0.1,
            }
        }

# Write to temporary file
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.pkl')
temp_filename = temp_file.name
temp_file.close()

with open(temp_filename, 'wb') as f:
    pickle.dump(eval_output, f)

ARGS = {
    'filename': temp_filename,
    'task': 'bbox',
    'metric': None,
    'prints': 'mPC',
    'aggregate': 'benchmark',
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (15, 6, 12) — shape (15, 6, 12) 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_results Ran
ran on a synthesized input
bethgelab/robust-detection-benchmark/data-analysis/robustness_eval.py
code served (permissive licence) · get_code("736f6acb4c7f3097")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (15, 6, 12) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import pickle
import tempfile
import os

# Create a temporary pickle file with mock evaluation output
# The structure needs to match what get_coco_style_results expects

# Create mock evaluation data
eval_output = {}

# Add 15 distortions (for benchmark aggregate mode)
distortion_names = [f'distortion_{i}' for i in range(15)]

for distortion in distortion_names:
    eval_output[distortion] = {}
    # Add 6 severity levels (0-5)
    for severity in range(6):
        eval_output[distortion][severity] = {
            'bbox': {
                'AP': 0.50 + np.random.rand() * 0.1,
                'AP50': 0.65 + np.random.rand() * 0.1,
                'AP75': 0.45 + np.random.rand() * 0.1,
                'APs': 0.30 + np.random.rand() * 0.1,
                'APm': 0.50 + np.random.rand() * 0.1,
                'APl': 0.65 + np.random.rand() * 0.1,
                'AR1': 0.55 + np.random.rand() * 0.1,
                'AR10': 0.60 + np.random.rand() * 0.1,
                'AR100': 0.62 + np.random.rand() * 0.1,
                'ARs': 0.40 + np.random.rand() * 0.1,
                'ARm': 0.60 + np.random.rand() * 0.1,
                'ARl': 0.70 + np.random.rand() * 0.1,
            }
        }

# Create temporary file
temp_fd, temp_path = tempfile.mkstemp(suffix='.pkl')
os.close(temp_fd)

try:
    with open(temp_path, 'wb') as f:
        pickle.dump(eval_output, f)
    
    ARGS = {
        'filename': temp_path,
        'task': 'bbox',
        'metric': None,
        'prints': 'mPC',
        'aggregate': 'benchmark'
    }
finally:
    # Note: In actual execution, you'd want to clean up temp_path after use
    pass
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (15, 6, 12) — shape (15, 6, 12) 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_voc_style_results Ran
ran on a synthesized input
bethgelab/robust-detection-benchmark/data-analysis/robustness_eval.py
code served (permissive licence) · get_code("d7b21a2a720fbd2b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (15, 6, 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 pickle
import numpy as np
import tempfile
import os

# Create a temporary pickle file with the expected data structure
# Structure: eval_output[distortion_name][severity_level][image_idx]['ap']
eval_output = {}

# Create 15 distortions for 'benchmark' mode (function uses [:15])
for distortion_idx in range(15):
    distortion_name = f'distortion_{distortion_idx}'
    eval_output[distortion_name] = {}
    
    # Create 6 severity levels (0-5)
    for severity in range(6):
        eval_output[distortion_name][severity] = []
        
        # Create 20 image results per severity level
        for img_idx in range(20):
            ap_value = np.random.rand() * 0.8 + 0.1  # AP values between 0.1 and 0.9
            eval_output[distortion_name][severity].append({'ap': float(ap_value)})

# Create a temporary file to store the pickle
temp_file = tempfile.NamedTemporaryFile(mode='wb', delete=False, suffix='.pkl')
pickle.dump(eval_output, temp_file)
temp_file.close()

ARGS = {
    'filename': temp_file.name,
    'prints': 'mPC',
    'aggregate': 'benchmark'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (15, 6, 1) — shape (15, 6, 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
plasma_fractal Ran
ran, honoured the contract we drafted
bethgelab/imagecorruptions/imagecorruptions/corruptions.py
code served (permissive licence) · get_code("33492980f6e79af5")
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 (256, 256) 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

ARGS = {
    "mapsize": 256,
    "wibbledecay": 3,
}
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 (256, 256) — shape (256, 256) 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

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 1 later) and 3 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
clipped_zoom 2 other papers
1 dated earlier, 1 later
1807.01697 (2018-07)
Benchmarking Neural Network Robustness to Common Corruptions and Surface Variations
2606.16996 (2026-06)
ActiveSAM: Image-Conditional Class Pruning for Fast and Accurate Open-Vocabulary Segmentation
plasma_fractal 1 other paper
1 dated earlier
1807.01697 (2018-07)
Benchmarking Neural Network Robustness to Common Corruptions and Surface Variations

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

The ability to detect objects regardless of image distortions or weather conditions is crucial for real-world applications of deep learning like autonomous driving. We here provide an easy-to-use benchmark to assess how object detection models perform when image quality degrades. The three resulting benchmark datasets, termed Pascal-C, Coco-C and Cityscapes-C, contain a large variety of image corruptions. We show that a range of standard object detection models suffer a severe performance loss on corrupted images (down to 30--60\% of the original performance). However, a simple data augmentation trick---stylizing the training images---leads to a substantial increase in robustness across corruption type, severity and dataset. We envision our comprehensive benchmark to track future progress towards building robust object detection models. Benchmark, code and data are publicly available.

For agents

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

get_harvested_code_for_paper("1907.07484")
get_code_for_paper("1907.07484")
have("1907.07484")

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

curl https://syntology.ai/api/ran/1907.07484.json

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

[![Syntology run record](https://syntology.ai/api/ran/1907.07484.svg)](https://syntology.ai/paper/1907.07484)

Connect an agent — have() is free.