SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.05894 · ICCV · 2021

MicroNet: Improving Image Recognition with Extremely Low FLOPs

Lei Zhang, Nuno Vasconcelos, Dongdong Chen, Lu Yuan, Zicheng Liu, Xiyang Dai, Yinpeng Chen, Mengchen Liu, Yunsheng Li, San Diego

arXiv · PDF · Open in the Atlas

Code that ran

10 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (6 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 7 have no contract verdict; 2 failures of our instrument; 1 no recorded run), measured 2026-11-22; 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/2108.05894.json · badge: /api/ran/2108.05894.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
liyunsheng13/micronet marked official in Papers with Code
Papers with Code archive
10 ran · 1 no recorded run
FunctionStatusWhere it lives
ChannelShuffle Ran
ran on a synthesized input · constructed an object; no result was computed
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("c678bfc0f2cc79c0")
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, ChannelShuffle: 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)

class ChannelShuffle(nn.Module):

    def __init__(self, groups):
        super(ChannelShuffle, self).__init__()
        self.groups = groups

    def forward(self, x):
        b, c, h, w = x.size()
        channels_per_group = c // self.groups
        x = x.view(b, self.groups, channels_per_group, h, w)
        x = torch.transpose(x, 1, 2).contiguous()
        out = x.view(b, -1, h, w)
        return out


# Create a valid input tensor
# batch=2, channels=4 (divisible by groups), height=3, width=5
x = torch.randn(2, 4, 3, 5)

# groups must divide the number of channels (4)
groups = 2

ARGS = {
    'groups': groups
}
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 ChannelShuffle — 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 ChannelShuffle 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
ChannelShuffle2 Ran
ran on a synthesized input · constructed an object; no result was computed
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("ff4b35233cab58ba")
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, ChannelShuffle2: 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)

class ChannelShuffle2(nn.Module):

    def __init__(self, groups):
        super(ChannelShuffle2, self).__init__()
        self.groups = groups

    def forward(self, x):
        b, c, h, w = x.size()
        channels_per_group = c // self.groups
        x = x.view(b, self.groups, channels_per_group, h, w)
        x = torch.transpose(x, 1, 2).contiguous()
        out = x.view(b, -1, h, w)
        return out

# Create a small test input tensor
# batch=2, channels=4, height=3, width=5
test_input = torch.randn(2, 4, 3, 5)

# groups must divide the number of channels evenly
groups = 2

ARGS = {
    "groups": groups
}
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 ChannelShuffle2 — 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 ChannelShuffle2 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
DYShiftMax Ran
ran on a synthesized input · constructed an object; no result was computed
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("4bdfb0e52e3dee51")
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, DYShiftMax: 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)

ARGS = {
    'inp': 16,
    'oup': 8,
    'reduction': 4,
    'act_max': 1.0,
    'act_relu': True,
    'init_a': [0.0, 0.0],
    'init_b': [0.0, 0.0],
    'relu_before_pool': False,
    'g': [1, 2],
    'expansion': 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 DYShiftMax — 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 DYShiftMax 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
DepthConv Ran
ran on a synthesized input
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("a6531130b3b20c72")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 4, 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 torch
import torch.nn as nn

class DepthConv(nn.Module):
    def __init__(self, inp, oup, kernel_size, stride):
        super(DepthConv, self).__init__()
        self.conv = nn.Sequential(nn.Conv2d(inp, oup, kernel_size, stride, kernel_size // 2, bias=False, groups=inp), nn.BatchNorm2d(oup))

    def forward(self, x):
        out = self.conv(x)
        return out

INIT_ARGS = {
    'inp': 4,
    'oup': 8,
    'kernel_size': 3,
    'stride': 1
}

# Construct input tensor: batch=2, channels=4, height=6, width=5
input_tensor = torch.randn(2, 4, 6, 5)

ARGS = {
    'x': input_tensor
}

# Build and call
model = DepthConv(**INIT_ARGS)
output = model(**ARGS)
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
DepthSpatialSepConv Ran
ran on a synthesized input · constructed an object; no result was computed
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("cd7aad73e9cd39f9")
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, DepthSpatialSepConv: 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)

ARGS = {
    'inp': 4,
    'expand': (2, 3),
    'kernel_size': 3,
    'stride': 1
}
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 DepthSpatialSepConv — 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 DepthSpatialSepConv 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
GroupConv Ran
ran on a synthesized input · constructed an object; no result was computed
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("07f1fa2646fb9c14")
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, GroupConv: 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)

ARGS = {
    'inp': 4,
    'oup': 8,
    'groups': [2]
}
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 GroupConv — 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 GroupConv 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
SELayer Ran
ran on a synthesized input · constructed an object; no result was computed
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("c4919e3470849b27")
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, SELayer: 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)

ARGS = {
    'inp': 16,
    'oup': 8,
    'reduction': 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 SELayer — 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 SELayer 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
get_act_layer Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("64775438a4130f56")
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 Sequential. 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)

class h_sigmoid(nn.Module):
    def __init__(self, inplace=True):
        super(h_sigmoid, self).__init__()
        self.relu = nn.ReLU6(inplace=inplace)

    def forward(self, x):
        return self.relu(x + 3) / 6

def _make_divisible(v, divisor, min_value=None):
    if min_value is None:
        min_value = divisor
    new_v = max(min_value, int(v + divisor / 2) // divisor * divisor)
    if new_v < 0.9 * v:
        new_v += divisor
    return new_v

def get_squeeze_channels(inp, reduction):
    if reduction == 4:
        squeeze = inp // reduction
    else:
        squeeze = _make_divisible(inp // reduction, 4)
    return squeeze

class SELayer(nn.Module):
    def __init__(self, inp, oup, reduction=4):
        super(SELayer, self).__init__()
        self.oup = oup
        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        squeeze = get_squeeze_channels(inp, reduction)
        self.fc = nn.Sequential(nn.Linear(inp, squeeze), nn.ReLU(inplace=True), nn.Linear(squeeze, oup), h_sigmoid())

    def forward(self, x):
        if isinstance(x, list):
            x_in = x[0]
            x_out = x[1]
        else:
            x_in = x
            x_out = x
        b, c, _, _ = x_in.size()
        y = self.avg_pool(x_in).view(b, c)
        y = self.fc(y).view(b, self.oup, 1, 1)
        return x_out * y

ARGS = {
    'inp': 16,
    'oup': 8,
    'mode': 'SE1',
    'act_relu': True,
    'act_max': 2,
    'act_bias': True,
    'init_a': [1.0, 0.0],
    'reduction': 4,
    'init_b': [0.0, 0.0],
    'g': None,
    'act': 'relu',
    'expansion': True
}
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 Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential output the recorded ndim 1 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
get_pointwise_conv Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("0ac4052c722dacb7")
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 GroupConv. 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)

ARGS = {
    'mode': 'group',
    'inp': 4,
    'oup': 8,
    'hiddendim': 16,
    'groups': [2]
}
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 GroupConv — a GroupConv output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a GroupConv 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
get_squeeze_channels Ran
ran, honoured the contract we drafted
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("5c6efc658367d250")
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 int, ndim 0, dtype family int, 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

def _make_divisible(v, divisor, min_value=None):
    """
    This function is taken from the original tf repo.
    It ensures that all layers have a channel number that is divisible by 8
    It can be seen here:
    https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py
    :param v:
    :param divisor:
    :param min_value:
    :return:
    """
    if min_value is None:
        min_value = divisor
    new_v = max(min_value, int(v + divisor / 2) // divisor * divisor)
    if new_v < 0.9 * v:
        new_v += divisor
    return new_v

def get_squeeze_channels(inp, reduction):
    if reduction == 4:
        squeeze = inp // reduction
    else:
        squeeze = _make_divisible(inp // reduction, 4)
    return squeeze

torch.manual_seed(0)

ARGS = {
    "inp": 64,
    "reduction": 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 int, ndim 0, dtype family int, 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
DYMicroBlock Not yet run
unverified (harvested, no recorded run)
liyunsheng13/micronet/backbone/micronet.py
code served (permissive licence) · get_code("78c65846facaa71e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 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 11 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 aims at addressing the problem of substantial performance degradation at extremely low computational cost (e.g. 5M FLOPs on ImageNet classification). We found that two factors, sparse connectivity and dynamic activation function, are effective to improve the accuracy. The former avoids the significant reduction of network width, while the latter mitigates the detriment of reduction in network depth. Technically, we propose micro-factorized convolution, which factorizes a convolution matrix into low rank matrices, to integrate sparse connectivity into convolution. We also present a new dynamic activation function, named Dynamic Shift Max, to improve the non-linearity via maxing out multiple dynamic fusions between an input feature map and its circular channel shift. Building upon these two new operators, we arrive at a family of networks, named MicroNet, that achieves significant performance gains over the state of the art in the low FLOP regime. For instance, under the constraint of 12M FLOPs, MicroNet achieves 59.4% top-1 accuracy on ImageNet classification, outperforming MobileNetV3 by 9.6%. Source code is at https://github.com/liyunsheng13/micronet.

For agents

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

get_harvested_code_for_paper("2108.05894")
get_code_for_paper("2108.05894")
have("2108.05894")

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

curl https://syntology.ai/api/ran/2108.05894.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.05894.svg)](https://syntology.ai/paper/2108.05894)

Connect an agent — have() is free.