SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2105.01883 · 2021

RepMLP: Re-parameterizing Convolutions into Fully-connected Layers for Image Recognition

arXiv · PDF · Open in the Atlas

Code that ran

10 of 14 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; 9 failures of our instrument; 4 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/2105.01883.json · badge: /api/ran/2105.01883.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
DingXiaoH/ResRep canonical (classifier label)
classifier basis: census
6 ran · 3 no recorded run
DingXiaoH/RepMLP canonical (classifier label)
classifier basis: classifier_2025
2 ran · 1 no recorded run
DingXiaoH/GSM-SGD canonical (classifier label)
classifier basis: census
2 ran · 0 no recorded run
FunctionStatusWhere it lives
calculate_mi1_flops Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
DingXiaoH/ResRep/rr/resrep_scripts.py
code served (permissive licence) · get_code("f480268edf8cc940")
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

# Analyze the function: calculate_mi1_flops takes a single parameter 'deps'
# deps must be a list/array of length 27
# Each element represents channel dependencies
# The function uses deps[layer_idx] and deps[-1]

# Create a valid deps array of length 27
# The assertion checks that len(deps) == 27 and that pairs match up
deps = np.array([
    64, 64,      # layer 0-1
    128, 128,    # layer 2-3
    256, 256,    # layer 4-5
    512, 512,    # layer 6-7
    512, 512,    # layer 8-9
    512, 512,    # layer 10-11
    1024, 1024,  # layer 12-13
    1024, 1024,  # layer 14-15
    1024, 1024,  # layer 16-17
    2048, 2048,  # layer 18-19
    2048, 2048,  # layer 20-21
    2048, 2048,  # layer 22-23
    2048, 2048,  # layer 24-25
    2048         # layer 26
], dtype=np.int64)

ARGS = {
    'deps': deps
}
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 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
conv_bn Ran
ran; our contract draft was wrong, not the code
DingXiaoH/RepMLP/repmlpnet.py
code served (permissive licence) · get_code("7186655759897af6")
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 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)

ARGS = {
    'in_channels': 3,
    'out_channels': 64,
    'kernel_size': 3,
    'stride': 1,
    'padding': 1,
    'groups': 1
}
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 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
conv_bn_relu Ran
ran; our contract draft was wrong, not the code
DingXiaoH/RepMLP/repmlpnet.py
code served (permissive licence) · get_code("18483b4d73618baa")
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)

ARGS = {
    'in_channels': 3,
    'out_channels': 64,
    'kernel_size': 3,
    'stride': 1,
    'padding': 1,
    'groups': 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 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
fuse_conv_bn Ran
ran; our contract draft was wrong, not the code
DingXiaoH/ResRep/rr/resrep_convert.py
code served (permissive licence) · get_code("ca447da2f48abbae")
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 numpy as np

# Create sample data for the save_dict
# The function expects conv and bn parameters with specific naming convention

# Create kernel (conv weight): shape (out_channels, in_channels, kernel_h, kernel_w)
kernel = np.random.randn(8, 3, 3, 3).astype(np.float32)

# Create batch norm parameters
running_mean = np.random.randn(8).astype(np.float32)
running_var = np.abs(np.random.randn(8)).astype(np.float32) + 0.1  # ensure positive
gamma = np.random.randn(8).astype(np.float32)
beta = np.random.randn(8).astype(np.float32)

# Create the save_dict with appropriate naming convention
save_dict = {
    'layer.conv.weight': kernel,
    'layer.bn.running_mean': running_mean,
    'layer.bn.running_var': running_var,
    'layer.bn.weight': gamma,
    'layer.bn.bias': beta,
}

# Create pop_name_set (will be modified in-place by the function)
pop_name_set = set()

# Define the kernel_name that will be used to construct other parameter names
kernel_name = 'layer.conv.weight'

# Construct ARGS dictionary
ARGS = {
    'save_dict': save_dict,
    'pop_name_set': pop_name_set,
    'kernel_name': kernel_name,
}
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 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
get_baseconfig_by_epoch Ran
ran; our contract draft was wrong, not the code
DingXiaoH/GSM-SGD/base_config.py
code served (permissive licence) · get_code("5bf521e2a90d2830")
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 BaseConfigByEpoch. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 numpy as np
from collections import namedtuple

torch.manual_seed(0)

ARGS = {
    'network_type': 'resnet50',
    'dataset_name': 'imagenet',
    'dataset_subset': 'train',
    'global_batch_size': 256,
    'num_node': 1,
    'weight_decay': 1e-4,
    'optimizer_type': 'sgd',
    'momentum': 0.9,
    'max_epochs': 100,
    'base_lr': 0.1,
    'lr_epoch_boundaries': [30, 60, 90],
    'lr_decay_factor': 0.1,
    'linear_final_lr': 0.0,
    'warmup_epochs': 5,
    'warmup_method': 'linear',
    'warmup_factor': 0.1,
    'ckpt_iter_period': 1000,
    'tb_iter_period': 100,
    'output_dir': '/tmp/output',
    'tb_dir': '/tmp/tb',
    'save_weights': True,
    'device': 'cpu',
    'weight_decay_bias': 0.0,
    'bias_lr_factor': 2.0,
    'init_weights': None,
    'val_epoch_period': 1,
    'grad_accum_iters': 1,
    'deps': None,
    'se_reduce_scale': 0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type BaseConfigByEpoch — a BaseConfigByEpoch 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
get_con_flops Ran
ran, honoured the contract we drafted
DingXiaoH/ResRep/rr/resrep_scripts.py
code served (permissive licence) · get_code("c6b81931ddcaf8b2")
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
import numpy as np

torch.manual_seed(0)

ARGS = {
    "input_deps": 4,
    "output_deps": 8,
    "h": 16,
    "w": 32,
    "kernel_size": 3,
    "groups": 2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, 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
get_cur_num_deactivated_filters Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
DingXiaoH/ResRep/rr/resrep_util.py
code served (permissive licence) · get_code("c615970236fd8e3c")
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 int64, 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 numpy as np

# Construct valid arguments for get_cur_num_deactivated_filters
origin_deps = np.array([10, 20, 15, 25], dtype=np.int64)
cur_deps = np.array([8, 15, 15, 20], dtype=np.int64)
follow_dict = {0: 0, 1: 2, 2: 2, 3: 3}

ARGS = {
    "origin_deps": origin_deps,
    "cur_deps": cur_deps,
    "follow_dict": follow_dict
}
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 int64, 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
load_cuda_data Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
DingXiaoH/GSM-SGD/gsm/gsm_train.py
code served (permissive licence) · get_code("dffbc817a4437a57")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 numpy as np

torch.manual_seed(0)

# Create a mock data loader for imagenet case
class ImageNetDataLoader:
    def __init__(self):
        self.batch_idx = 0
    
    def __iter__(self):
        return self
    
    def __next__(self):
        if self.batch_idx >= 1:
            raise StopIteration
        self.batch_idx += 1
        data = np.random.randn(2, 4).astype(np.float32)
        label = np.array([0, 1], dtype=np.int64)
        return {'data': data, 'label': label}

# Create a mock data loader for non-imagenet case
class StandardDataLoader:
    def __init__(self):
        self.batch_idx = 0
    
    def __iter__(self):
        return self
    
    def __next__(self):
        if self.batch_idx >= 1:
            raise StopIteration
        self.batch_idx += 1
        data = torch.randn(2, 4)
        label = torch.tensor([0, 1], dtype=torch.long)
        return (data, label)

# Test with imagenet dataset
imagenet_loader = ImageNetDataLoader()
ARGS = {
    'data_loader': imagenet_loader,
    'dataset_name': 'imagenet'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, 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
resrep_get_deps_and_metric_dict Ran
ran; our contract draft was wrong, not the code
DingXiaoH/ResRep/rr/resrep_util.py
code served (permissive licence) · get_code("854695e3e20a0fff")
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
import torch.nn as nn
import numpy as np

torch.manual_seed(0)

# Create a custom module class that has the required attributes and methods
class MockConvModule(nn.Module):
    def __init__(self, conv_idx, num_mask_ones, metric_vector_size):
        super().__init__()
        self.conv_idx = conv_idx
        self._num_mask_ones = num_mask_ones
        self._metric_vector = np.random.rand(metric_vector_size).tolist()
    
    def get_num_mask_ones(self):
        return self._num_mask_ones
    
    def get_metric_vector(self):
        return self._metric_vector

# Create a model with child modules that have conv_idx
class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = MockConvModule(conv_idx=0, num_mask_ones=3, metric_vector_size=2)
        self.conv2 = MockConvModule(conv_idx=1, num_mask_ones=2, metric_vector_size=2)
        self.conv3 = MockConvModule(conv_idx=2, num_mask_ones=4, metric_vector_size=2)
    
model = MockModel()

# Create origin_deps array - must have at least as many mask ones as the model layers
origin_deps = np.array([5, 3, 6], dtype=int)

# Create pacesetter_dict - maps follower indices to pacesetter indices
pacesetter_dict = {
    1: 0,  # follower 1 has pacesetter 0
    2: 1,  # follower 2 has pacesetter 1
}

ARGS = {
    "origin_deps": origin_deps,
    "model": model,
    "pacesetter_dict": pacesetter_dict,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, 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
resrep_get_layer_mask_ones_and_metric_dict Ran
ran; our contract draft was wrong, not the code
DingXiaoH/ResRep/rr/resrep_util.py
code served (permissive licence) · get_code("2249cf3b5f3510fd")
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, ndim 1, 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

torch.manual_seed(0)

class MockConvModule(nn.Module):
    def __init__(self, conv_idx, num_mask_ones, metric_vector_size):
        super().__init__()
        self.conv_idx = conv_idx
        self._num_mask_ones = num_mask_ones
        self._metric_vector = torch.randn(metric_vector_size)
        self.conv = nn.Conv2d(3, 16, 3)
    
    def get_num_mask_ones(self):
        return self._num_mask_ones
    
    def get_metric_vector(self):
        return self._metric_vector

class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv_layer_1 = MockConvModule(conv_idx=0, num_mask_ones=12, metric_vector_size=5)
        self.conv_layer_2 = MockConvModule(conv_idx=1, num_mask_ones=8, metric_vector_size=5)
        self.linear = nn.Linear(10, 2)

model = MockModel()

ARGS = {
    "model": model
}
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, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
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
calculate_resnet_bottleneck_flops Not yet run
unverified (harvested, no recorded run)
DingXiaoH/ResRep/rr/resrep_scripts.py
code served (permissive licence) · get_code("4f77c99e445c50d0")
How it ran

no invocation recorded.

fold_conv Not yet run
unverified (harvested, no recorded run)
DingXiaoH/ResRep/rr/resrep_convert.py
code served (permissive licence) · get_code("683ccc67b0312bd0")
How it ran

no invocation recorded.

fuse_bn Not yet run
unverified (harvested, no recorded run)
DingXiaoH/RepMLP/repmlpnet.py
code served (permissive licence) · get_code("c1e5b17d52a4eaa8")
How it ran

no invocation recorded.

get_baseconfig_by_epoch Not yet run
unverified (harvested, no recorded run)
DingXiaoH/ResRep/base_config.py
code served (permissive licence) · get_code("15826dfdec9eddac")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 4 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 3 later) and 10 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
conv_bn 10 other papers
2 dated earlier, 8 later
2003.13678 (2020-03)
Designing Network Design Spaces
2101.03697 (2021-01)
RepVGG: Making VGG-style ConvNets Great Again
2112.11081 (2021-12)
RepMLPNet: Hierarchical Vision MLP with Re-parameterized Locality
and 7 more not named here
conv_bn_relu 1 other paper
1 later
2112.11081 (2021-12)
RepMLPNet: Hierarchical Vision MLP with Re-parameterized Locality
fuse_bn 1 other paper
1 later
2112.11081 (2021-12)
RepMLPNet: Hierarchical Vision MLP with Re-parameterized Locality
get_baseconfig_by_epoch 1 other paper
1 dated earlier
1909.12778 (2019-09)
Global Sparse Momentum SGD for Pruning Very Deep Neural Networks

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

We propose RepMLP, a multi-layer-perceptron-style neural network building block for image recognition, which is composed of a series of fully-connected (FC) layers. Compared to convolutional layers, FC layers are more efficient, better at modeling the long-range dependencies and positional patterns, but worse at capturing the local structures, hence usually less favored for image recognition. We propose a structural re-parameterization technique that adds local prior into an FC to make it powerful for image recognition. Specifically, we construct convolutional layers inside a RepMLP during training and merge them into the FC for inference. On CIFAR, a simple pure-MLP model shows performance very close to CNN. By inserting RepMLP in traditional CNN, we improve ResNets by 1.8% accuracy on ImageNet, 2.9% for face recognition, and 2.3% mIoU on Cityscapes with lower FLOPs. Our intriguing findings highlight that combining the global representational capacity and positional perception of FC with the local prior of convolution can improve the performance of neural network with faster speed on both the tasks with translation invariance (e.g., semantic segmentation) and those with aligned images and positional patterns (e.g., face recognition). The code and models are available at https://github.com/DingXiaoH/RepMLP.

For agents

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

get_harvested_code_for_paper("2105.01883")
get_code_for_paper("2105.01883")
have("2105.01883")

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

curl https://syntology.ai/api/ran/2105.01883.json

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

[![Syntology run record](https://syntology.ai/api/ran/2105.01883.svg)](https://syntology.ai/paper/2105.01883)

Connect an agent — have() is free.