SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2103.06255 · CVPR · 2021

Involution: Inverting the Inherence of Convolution for Visual Recognition

Qi She, Tong Zhang, Qifeng Chen, Changhu Wang, Xiangtai Li, Jie Hu, Lei Zhu, Duo Li

arXiv · PDF · Open in the Atlas

Code that ran

7 of 9 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (7 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 7 have no contract verdict; 0 failures of our instrument; 2 no recorded run), measured 2026-11-21; 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/2103.06255.json · badge: /api/ran/2103.06255.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
shikishima-TasakiLab/Involution-PyTorch community
Papers with Code archive
1 ran · 1 no recorded run
YashovardhanSrivastava19/Inv_Conv_Model community
Papers with Code archive
1 ran · 1 no recorded run
shuuchen/involution.pytorch community
Papers with Code archive
1 ran · 0 no recorded run
xmu-xiaoma666/External-Attention-pytorch community
Papers with Code archive
1 ran · 0 no recorded run
YirunKCL/Tensorflow-Keras-Involution2D community
Papers with Code archive
1 ran · 0 no recorded run
PrivateMaRyan/keras-involution2Ds community
Papers with Code archive
1 ran · 0 no recorded run
ChristophReich1996/Involution community
Papers with Code archive
1 ran · 0 no recorded run
FunctionStatusWhere it lives
Involution Ran
ran on a synthesized input · constructed an object; no result was computed
shuuchen/involution.pytorch/involution.py
code served (permissive licence) · get_code("2ab588f06b5fcaf4")
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, Involution: 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)

# Involution expects:
# - in_channels: int
# - out_channels: int
# - groups: int (default 1)
# - kernel_size: int (default 3)
# - stride: int (default 1)
# - reduction_ratio: int (default 2)

ARGS = {
    'in_channels': 4,
    'out_channels': 8,
    'groups': 1,
    'kernel_size': 3,
    'stride': 1,
    'reduction_ratio': 2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Involution — 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 Involution 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
Involution Ran
ran on a synthesized input · constructed an object; no result was computed
xmu-xiaoma666/External-Attention-pytorch/model/conv/Involution.py
code served (permissive licence) · get_code("7d4e496bf8ce2354")
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, Involution: 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
import torch.nn as nn

torch.manual_seed(0)

# Create a minimal input tensor for testing
# Shape: (batch=2, channels=4, height=8, width=8)
test_input = torch.randn(2, 4, 8, 8)

# Define ARGS dictionary with all parameters for Involution.__init__
ARGS = {
    'kernel_size': 3,
    'in_channel': 4,
    'stride': 1,
    'group': 1,
    'ratio': 4
}
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 Involution — 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 Involution 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
Involution2D Ran
ran on a synthesized input · constructed an object; no result was computed
YirunKCL/Tensorflow-Keras-Involution2D/involution2d.py
pointer only (licence: NONE) · get_code("fab2bfdc9b7802ab")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, YirunKCL/Tensorflow-Keras-Involution2D 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.

Involution2D Ran
ran on a synthesized input · constructed an object; no result was computed
PrivateMaRyan/keras-involution2Ds/Involution2D.py
pointer only (licence: NONE) · get_code("b4e4c359cd255f73")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, PrivateMaRyan/keras-involution2Ds 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.

Involution2d Ran
ran on a synthesized input · constructed an object; no result was computed
ChristophReich1996/Involution/involution/involution.py
code served (permissive licence) · get_code("56639e9724c537c5")
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, Involution2d: 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
import torch.nn as nn

torch.manual_seed(0)

# Define a custom sigma_mapping module for testing
class CustomSigmaMapping(nn.Module):
    def __init__(self, channels):
        super().__init__()
        self.bn = nn.BatchNorm2d(channels, momentum=0.3)
        self.relu = nn.ReLU()
    
    def forward(self, x):
        return self.relu(self.bn(x))

ARGS = {
    'in_channels': 16,
    'out_channels': 32,
    'sigma_mapping': None,  # Will use default BN + ReLU
    'kernel_size': (7, 7),
    'stride': (1, 1),
    'groups': 4,
    'reduce_ratio': 2,
    'dilation': (1, 1),
    'padding': (3, 3),
    'bias': False,
    'force_shape_match': 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 Involution2d — 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 Involution2d 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
Involution2d Ran
ran on a synthesized input · constructed an object; no result was computed
shikishima-TasakiLab/Involution-PyTorch/involution/involution2d.py
pointer only (licence: NOASSERTION) · get_code("0fddcb87a4ff3ec5")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, shikishima-TasakiLab/Involution-PyTorch (NOASSERTION). 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.

InvolutionLayer Ran
ran on a synthesized input · constructed an object; no result was computed
YashovardhanSrivastava19/Inv_Conv_Model/InConvModel.py
code served (permissive licence) · get_code("c541633190c973cd")
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, InvolutionLayer: 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 tensorflow as tf
import numpy as np

# Set seeds for reproducibility
np.random.seed(0)
tf.random.set_seed(0)

# Create a small sample input tensor
# Shape: (batch_size, height, width, channels)
sample_input = tf.constant(
    np.random.randn(2, 8, 8, 16).astype(np.float32),
    dtype=tf.float32
)

# Define arguments for InvolutionLayer
ARGS = {
    'Channel': 16,
    'Group_Number': 4,
    'Kernel_Size': 3,
    'Stride': 1,
    'Reduction_Ratio': 4
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type InvolutionLayer — 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 InvolutionLayer 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
Configuration Not yet run
unverified (harvested, no recorded run)
YashovardhanSrivastava19/Inv_Conv_Model/InConvModel.py
code served (permissive licence) · get_code("53278de7fe970d81")
How it ran

no invocation recorded.

_involution2d Not yet run
unverified (harvested, no recorded run)
shikishima-TasakiLab/Involution-PyTorch/involution/involution2d.py
pointer only (licence: NOASSERTION) · get_code("15063abf3990f72c")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, shikishima-TasakiLab/Involution-PyTorch (NOASSERTION). 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 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 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 9 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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Convolution has been the core ingredient of modern neural networks, triggering the surge of deep learning in vision. In this work, we rethink the inherent principles of standard convolution for vision tasks, specifically spatialagnostic and channel-specific. Instead, we present a novel atomic operation for deep neural networks by inverting the aforementioned design principles of convolution, coined as involution. We additionally demystify the recent popular self-attention operator and subsume it into our involution family as an over-complicated instantiation. The proposed involution operator could be leveraged as fundamental bricks to build the new generation of neural networks for visual recognition, powering different deep learning models on several prevalent benchmarks, including Im-ageNet classification, COCO detection and segmentation, together with Cityscapes segmentation. Our involutionbased models improve the performance of convolutional baselines using ResNet-50 by up to 1.6% top-1 accuracy, 2.5% and 2.4% bounding box AP, and 4.7% mean IoU absolutely while compressing the computational cost to 66%, 65%, 72%, and 57% on the above benchmarks, respectively. Code and pre-trained models for all the tasks are available at https://github.com/d-li14/involution.

For agents

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

get_harvested_code_for_paper("2103.06255")
get_code_for_paper("2103.06255")
have("2103.06255")

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

curl https://syntology.ai/api/ran/2103.06255.json

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

[![Syntology run record](https://syntology.ai/api/ran/2103.06255.svg)](https://syntology.ai/paper/2103.06255)

Connect an agent — have() is free.