SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2008.06822 · 2020

Relevance Attack on Detectors

arXiv · PDF · Open in the Atlas

Code that ran

4 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; 0 honoured the contract we drafted, 0 violated it and 4 have no contract verdict; 0 failures of our instrument; 1 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/2008.06822.json · badge: /api/ran/2008.06822.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
allenchen1998/rad canonical (classifier label)
classifier basis: census
4 ran · 1 no recorded run
FunctionStatusWhere it lives
attack_classification_model Ran
ran on a synthesized input
allenchen1998/rad/retinanet_base.py
code served (permissive licence) · get_code("38aca89469ae28f6")
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 function. 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 keras
import tensorflow as tf
import numpy as np

# Reset default graph to ensure clean state
tf.compat.v1.reset_default_graph()

ARGS = {
    'num_classes': 10,
    'num_anchors': 9,
    'pyramid_feature_size': 256,
    'prior_probability': 0.01,
    'classification_feature_size': 256,
    'name': 'classification_submodel'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
keras, numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
calculate_direction Ran
ran on a synthesized input
allenchen1998/rad/utils.py
code served (permissive licence) · get_code("15cf4eb56d4f1c6c")
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 float32, 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 numpy as np
import cv2

# Define a simple mock function for run_direction
# It should accept an image array and return a scalar value
def mock_run_direction(adv_image):
    """Mock function that takes an image and returns a scalar directional value."""
    return np.sum(adv_image) * 0.001  # Return a scalar

# Create a sample adversarial image
# Shape: (batch, height, width, channels)
# Using distinct sizes: batch=1, height=4, width=3, channels=2
sample_adv_image = np.random.randn(1, 4, 3, 2).astype(np.float32)

# Define ARGS dictionary with all parameters for calculate_direction
ARGS = {
    'adv_image': sample_adv_image,
    'run_direction': mock_run_direction,
    'DI': False,
    'SI': False
}
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 float32, 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
create_pyramid_features_attack Ran
ran on a synthesized input
allenchen1998/rad/retinanet_base.py
code served (permissive licence) · get_code("0074253cf9e6cb74")
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 keras
import keras.layers
import numpy as np

# Create small feature maps as Keras inputs
# C3, C4, C5 are feature maps from a ResNet backbone
# We need to create them as Keras tensors/layers

# Create input layers with small spatial dimensions
# Batch=2, channels vary, spatial dims decrease
C3_input = keras.layers.Input(shape=(16, 16, 64), name='C3_input')
C4_input = keras.layers.Input(shape=(8, 8, 128), name='C4_input')
C5_input = keras.layers.Input(shape=(4, 4, 256), name='C5_input')

# These are Keras tensors that can be used as inputs to the function
C3 = C3_input
C4 = C4_input
C5 = C5_input

ARGS = {
    'C3': C3,
    'C4': C4,
    'C5': C5,
    'feature_size': 256
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
keras, 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
random_padding Ran
ran on a synthesized input
allenchen1998/rad/utils.py
code served (permissive licence) · get_code("98695980414840db")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 4, dtype family float, shape (1, 8, 12, 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

# Construct a valid adversarial image tensor
# The function expects adv_image to have shape (batch, height, width, channels)
# Based on the function body:
# - adv_image.shape[0] is batch dimension
# - adv_image.shape[1] is height
# - adv_image.shape[2] is width
# - adv_image.shape[3] is channels (implied by the slicing)

np.random.seed(0)

# Create a small test image with shape (1, 8, 12, 3) for RGB image
# Using distinct dimensions: batch=1, height=8, width=12, channels=3
adv_image = np.random.rand(1, 8, 12, 3).astype(np.float32) * 255

ARGS = {
    "adv_image": adv_image
}
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 4, dtype family float, shape (1, 8, 12, 3) — shape (1, 8, 12, 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
build_direction Not yet run
unverified (harvested, no recorded run)
allenchen1998/rad/utils.py
code served (permissive licence) · get_code("007be9a56731eebc")
How it ran

no invocation recorded.

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, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 5 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.

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

This paper focuses on high-transferable adversarial attacks on detectors, which are hard to attack in a black-box manner, because of their multiple-output characteristics and the diversity across architectures. To pursue a high attack transferability, one plausible way is to find a common property across detectors, which facilitates the discovery of common weaknesses. We are the first to suggest that the relevance map from interpreters for detectors is such a property. Based on it, we design a Relevance Attack on Detectors (RAD), which achieves a state-of-the-art transferability, exceeding existing results by above 20%. On MS COCO, the detection mAPs for all 8 black-box architectures are more than halved and the segmentation mAPs are also significantly influenced. Given the great transferability of RAD, we generate the first adversarial dataset for object detection and instance segmentation, i.e., Adversarial Objects in COntext (AOCO), which helps to quickly evaluate and improve the robustness of detectors.

For agents

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

get_harvested_code_for_paper("2008.06822")
get_code_for_paper("2008.06822")
have("2008.06822")

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

curl https://syntology.ai/api/ran/2008.06822.json

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

[![Syntology run record](https://syntology.ai/api/ran/2008.06822.svg)](https://syntology.ai/paper/2008.06822)

Connect an agent — have() is free.