SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2103.12579 · CVPR · 2021

MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognition

Shuang Li, Yulin Wang, Kaixiong Gong, Chi Liu, Xinjing Cheng, Feng Qiao

arXiv · PDF · Open in the Atlas

Code that ran

2 of 3 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 2 have no contract verdict; 0 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/2103.12579.json · badge: /api/ran/2103.12579.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
BIT-DA/MetaSAug marked official in Papers with Code
Papers with Code archive
2 ran · 1 no recorded run
FunctionStatusWhere it lives
MetaBatchNorm2d Ran
ran on a synthesized input
BIT-DA/MetaSAug/ImageNet_iNat/resnet_meta.py
code served (permissive licence) · get_code("885f4cef722f0ecb")
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. 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 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
import torch.nn.functional as F
from torch.autograd import Variable

def to_var(x, requires_grad=True):
    return Variable(x, requires_grad=requires_grad)

class MetaModule(nn.Module):

    def params(self):
        for name, param in self.named_params(self):
            yield param

    def named_leaves(self):
        return []

    def named_submodules(self):
        return []

    def named_params(self, curr_module=None, memo=None, prefix=''):
        if memo is None:
            memo = set()
        if hasattr(curr_module, 'named_leaves'):
            for name, p in curr_module.named_leaves():
                if p is not None and p not in memo:
                    memo.add(p)
                    yield (prefix + ('.' if prefix else '') + name, p)
        else:
            for name, p in curr_module._parameters.items():
                if p is not None and p not in memo:
                    memo.add(p)
                    yield (prefix + ('.' if prefix else '') + name, p)
        for mname, module in curr_module.named_children():
            submodule_prefix = prefix + ('.' if prefix else '') + mname
            for name, p in self.named_params(module, memo, submodule_prefix):
                yield (name, p)

    def update_params(self, lr_inner, first_order=False, source_params=None, detach=False):
        if source_params is not None:
            for tgt, src in zip(self.named_params(self), source_params):
                name_t, param_t = tgt
                grad = src
                if first_order:
                    grad = to_var(grad.detach().data)
                tmp = param_t - lr_inner * grad
                self.set_param(self, name_t, tmp)
        else:
            for name, param in self.named_params(self):
                if not detach:
                    grad = param.grad
                    if first_order:
                        grad = to_var(grad.detach().data)
                    tmp = param - lr_inner * grad
                    self.set_param(self, name, tmp)
                else:
                    param = param.detach_()
                    self.set_param(self, name, param)

    def set_param(self, curr_mod, name, param):
        if '.' in name:
            n = name.split('.')
            module_name = n[0]
            rest = '.'.join(n[1:])
            for name, mod in curr_mod.named_children():
                if module_name == name:
                    self.set_param(mod, rest, param)
                    break
        else:
            setattr(curr_mod, name, param)

    def detach_params(self):
        for name, param in self.named_params(self):
            self.set_param(self, name, param.detach())

    def copy(self, other, same_var=False):
        for name, param in other.named_params():
            if not same_var:
                param = to_var(param.data.clone(), requires_grad=True)
            self.set_param(name, param)

class MetaBatchNorm2d(MetaModule):

    def __init__(self, *args, **kwargs):
        super().__init__()
        ignore = nn.BatchNorm2d(*args, **kwargs)
        self.num_features = ignore.num_features
        self.eps = ignore.eps
        self.momentum = ignore.momentum
        self.affine = ignore.affine
        self.track_running_stats = ignore.track_running_stats
        if self.affine:
            self.register_buffer('weight', to_var(ignore.weight.data, requires_grad=True))
            self.register_buffer('bias', to_var(ignore.bias.data, requires_grad=True))
        if self.track_running_stats:
            self.register_buffer('running_mean', torch.zeros(self.num_features))
            self.register_buffer('running_var', torch.ones(self.num_features))
        else:
            self.register_parameter('running_mean', None)
            self.register_parameter('running_var', None)

    def forward(self, x):
        return F.batch_norm(x, self.running_mean, self.running_var, self.weight, self.bias, self.training or not self.track_running_stats, self.momentum, self.eps)

    def named_leaves(self):
        return [('weight', self.weight), ('bias', self.bias)]


INIT_ARGS = {
    'num_features': 4,
}

ARGS = {
    'x': torch.randn(2, 4, 3, 3),
}
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)
MetaConv2d Ran
ran on a synthesized input · constructed an object; no result was computed
BIT-DA/MetaSAug/ImageNet_iNat/resnet_meta.py
code served (permissive licence) · get_code("c5563b3215911fe1")
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, MetaConv2d: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn
from torch.autograd import Variable

torch.manual_seed(0)

# Define ARGS dict with all parameters for MetaConv2d.__init__
ARGS = {
    'in_channels': 3,
    'out_channels': 4,
    'kernel_size': 3,
    'stride': 1,
    'padding': 1,
    'dilation': 1,
    'groups': 1,
    'bias': True,
    'padding_mode': 'zeros',
    'device': None,
    'dtype': None
}
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 MetaConv2d — 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 MetaConv2d 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
FeatureMeta Not yet run
unverified (harvested, no recorded run)
BIT-DA/MetaSAug/ImageNet_iNat/resnet_meta.py
code served (permissive licence) · get_code("9656bad1da39d4b0")
How it ran

no invocation recorded.

Also present in other papers' code

Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 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 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.

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

To this end, we propose a novel approach to learn transformed semantic directions with metalearning automatically. In specific, the augmentation strategy during training is dynamically optimized, aiming to minimize the loss on a small balanced validation set, which is approximated via a meta update step. Extensive empirical results on CIFAR-LT-10/100, ImageNet-LT, and iNaturalist 2017/2018 validate the effectiveness of our method.

For agents

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

get_harvested_code_for_paper("2103.12579")
get_code_for_paper("2103.12579")
have("2103.12579")

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

curl https://syntology.ai/api/ran/2103.12579.json

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

[![Syntology run record](https://syntology.ai/api/ran/2103.12579.svg)](https://syntology.ai/paper/2103.12579)

Connect an agent — have() is free.