SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.02688 · CVPR · 2022

Zoom In and Out: A Mixed-scale Triplet Network for Camouflaged Object Detection

Xiaoqi Zhao, Youwei Pang, Lihe Zhang, Huchuan Lu, Tian-Zhu Xiang

arXiv · PDF · Open in the Atlas

Code that ran

12 of 13 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; 1 honoured the contract we drafted, 0 violated it and 7 have no contract verdict; 4 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/2203.02688.json · badge: /api/ran/2203.02688.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
lartpang/ZoomNet community
Papers with Code archive
12 ran · 1 no recorded run
FunctionStatusWhere it lives
ASPP Ran
ran on a synthesized input · constructed an object; no result was computed
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("baaa14215b7042ce")
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, ASPP: 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
from timm.models.layers import to_2tuple

torch.manual_seed(0)

def _get_act_fn(act_name, inplace=True):
    if act_name == 'relu':
        return nn.ReLU(inplace=inplace)
    elif act_name == 'leaklyrelu':
        return nn.LeakyReLU(negative_slope=0.1, inplace=inplace)
    elif act_name == 'gelu':
        return nn.GELU()
    else:
        raise NotImplementedError

class ConvBNReLU(nn.Sequential):
    def __init__(self, in_planes, out_planes, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=False, act_name='relu', is_transposed=False):
        super().__init__()
        if is_transposed:
            conv_module = nn.ConvTranspose2d
        else:
            conv_module = nn.Conv2d
        self.add_module(name='conv', module=conv_module(in_planes, out_planes, kernel_size=kernel_size, stride=to_2tuple(stride), padding=to_2tuple(padding), dilation=to_2tuple(dilation), groups=groups, bias=bias))
        self.add_module(name='bn', module=nn.BatchNorm2d(out_planes))
        if act_name is not None:
            self.add_module(name=act_name, module=_get_act_fn(act_name=act_name))

def cus_sample(x, mode='size', factors=None):
    """Stand-in for custom sampling function used in ASPP.forward"""
    if mode == 'size' and factors is not None:
        return torch.nn.functional.interpolate(x, size=factors, mode='nearest')
    return x

class ASPP(nn.Module):
    def __init__(self, in_dim, out_dim):
        super(ASPP, self).__init__()
        self.conv1 = ConvBNReLU(in_dim, out_dim, kernel_size=1)
        self.conv2 = ConvBNReLU(in_dim, out_dim, kernel_size=3, dilation=2, padding=2)
        self.conv3 = ConvBNReLU(in_dim, out_dim, kernel_size=3, dilation=5, padding=5)
        self.conv4 = ConvBNReLU(in_dim, out_dim, kernel_size=3, dilation=7, padding=7)
        self.conv5 = ConvBNReLU(in_dim, out_dim, kernel_size=1)
        self.fuse = ConvBNReLU(5 * out_dim, out_dim, 3, 1, 1)

    def forward(self, x):
        conv1 = self.conv1(x)
        conv2 = self.conv2(x)
        conv3 = self.conv3(x)
        conv4 = self.conv4(x)
        conv5 = self.conv5(cus_sample(x.mean((2, 3), keepdim=True), mode='size', factors=x.size()[2:]))
        return self.fuse(torch.cat((conv1, conv2, conv3, conv4, conv5), 1))

ARGS = {
    'in_dim': 4,
    'out_dim': 3,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ASPP — 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 ASPP 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
BasicModelClass Ran
ran on a synthesized input · constructed an object; no result was computed
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("05985573b34cdbec")
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, BasicModelClass: 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
from torch.nn.functional import binary_cross_entropy_with_logits

torch.manual_seed(0)

# Create a concrete implementation of BasicModelClass for testing
class ConcreteModel(nn.Module):
    def __init__(self):
        super(ConcreteModel, self).__init__()
        self.is_training = True
        self.linear = nn.Linear(4, 3)

    def forward(self, *args, **kwargs):
        if not self.is_training or not self.training:
            results = self.test_forward(*args, **kwargs)
        else:
            results = self.train_forward(*args, **kwargs)
        return results

    def train_forward(self, x, y=None):
        return self.linear(x)

    def test_forward(self, x, y=None):
        return self.linear(x)

    def cal_loss(self, preds, targets):
        loss = binary_cross_entropy_with_logits(
            input=preds, 
            target=targets, 
            reduction="mean"
        )
        loss_str = f"loss:{loss.item():.5f}"
        return loss, loss_str


# Create sample data
batch_size = 2
input_dim = 4
output_dim = 3

sample_input = torch.randn(batch_size, input_dim)
sample_target = torch.randint(0, 2, (batch_size, output_dim)).float()

# Create model instance
model = ConcreteModel()
model.train()

# Define ARGS dictionary with parameters for BasicModelClass.__init__
# BasicModelClass.__init__ takes no parameters besides self
ARGS = {}
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 BasicModelClass — 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 BasicModelClass 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
ConvBNReLU Ran
ran on a synthesized input · constructed an object; no result was computed
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("862d137d282a2bab")
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, ConvBNReLU: 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
from timm.models.layers import to_2tuple

torch.manual_seed(0)

ARGS = {
    'in_planes': 3,
    'out_planes': 16,
    'kernel_size': 3,
    'stride': 1,
    'padding': 1,
    'dilation': 1,
    'groups': 1,
    'bias': False,
    'act_name': 'relu',
    'is_transposed': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ConvBNReLU — 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 ConvBNReLU 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
HMU Ran
ran on a synthesized input · constructed an object; no result was computed
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("7643f062b521c4d3")
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, HMU: 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)

ARGS = {
    'in_c': 16,
    'num_groups': 4,
    'hidden_dim': 8,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type HMU — 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 HMU 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
Registry Ran
ran on a synthesized input · constructed an object; no result was computed
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("4a417a63c358c5ab")
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, Registry: 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
from typing import Any, Dict, Iterable, Iterator, Tuple

torch.manual_seed(0)

# Since Registry.__init__ only takes a 'name' parameter
ARGS = {
    'name': 'test_registry'
}
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 Registry — the run constructed an instance of its own class and did not call it; no shape 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
SIU Ran
ran on a synthesized input · constructed an object; no result was computed
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("aec31add8468eb86")
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, SIU: 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
from torch.nn import functional as F

torch.manual_seed(0)

def cus_sample(x, mode='size', factors=None):
    """Stand-in for the cus_sample function used in forward"""
    if mode == 'size' and factors is not None:
        return F.interpolate(x, size=factors, mode='bilinear', align_corners=False)
    return x

def _get_act_fn(act_name, inplace=True):
    if act_name == 'relu':
        return nn.ReLU(inplace=inplace)
    elif act_name == 'leaklyrelu':
        return nn.LeakyReLU(negative_slope=0.1, inplace=inplace)
    elif act_name == 'gelu':
        return nn.GELU()
    else:
        raise NotImplementedError

class ConvBNReLU(nn.Sequential):
    def __init__(self, in_planes, out_planes, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=False, act_name='relu', is_transposed=False):
        super().__init__()
        if is_transposed:
            conv_module = nn.ConvTranspose2d
        else:
            conv_module = nn.Conv2d
        
        from timm.models.layers import to_2tuple
        self.add_module(name='conv', module=conv_module(in_planes, out_planes, kernel_size=kernel_size, stride=to_2tuple(stride), padding=to_2tuple(padding), dilation=to_2tuple(dilation), groups=groups, bias=bias))
        self.add_module(name='bn', module=nn.BatchNorm2d(out_planes))
        if act_name is not None:
            self.add_module(name=act_name, module=_get_act_fn(act_name=act_name))

class SIU(nn.Module):
    def __init__(self, in_dim):
        super().__init__()
        self.conv_l_pre_down = ConvBNReLU(in_dim, in_dim, 5, stride=1, padding=2)
        self.conv_l_post_down = ConvBNReLU(in_dim, in_dim, 3, 1, 1)
        self.conv_m = ConvBNReLU(in_dim, in_dim, 3, 1, 1)
        self.conv_s_pre_up = ConvBNReLU(in_dim, in_dim, 3, 1, 1)
        self.conv_s_post_up = ConvBNReLU(in_dim, in_dim, 3, 1, 1)
        self.trans = nn.Sequential(ConvBNReLU(3 * in_dim, in_dim, 1), ConvBNReLU(in_dim, in_dim, 3, 1, 1), ConvBNReLU(in_dim, in_dim, 3, 1, 1), nn.Conv2d(in_dim, 3, 1))

    def forward(self, l, m, s, return_feats=False):
        tgt_size = m.shape[2:]
        l = self.conv_l_pre_down(l)
        l = F.adaptive_max_pool2d(l, tgt_size) + F.adaptive_avg_pool2d(l, tgt_size)
        l = self.conv_l_post_down(l)
        m = self.conv_m(m)
        s = self.conv_s_pre_up(s)
        s = cus_sample(s, mode='size', factors=m.shape[2:])
        s = self.conv_s_post_up(s)
        attn = self.trans(torch.cat([l, m, s], dim=1))
        attn_l, attn_m, attn_s = torch.softmax(attn, dim=1).chunk(3, dim=1)
        lms = attn_l * l + attn_m * m + attn_s * s
        if return_feats:
            return (lms, dict(attn_l=attn_l, attn_m=attn_m, attn_s=attn_s, l=l, m=m, s=s))
        return lms

in_dim = 4
siu_model = SIU(in_dim=in_dim)

batch_size = 2
height = 8
width = 8

l_tensor = torch.randn(batch_size, in_dim, height, width)
m_tensor = torch.randn(batch_size, in_dim, height, width)
s_tensor = torch.randn(batch_size, in_dim, height, width)

ARGS = {
    'in_dim': in_dim,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SIU — 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 SIU 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
TransLayer Ran
ran on a synthesized input · constructed an object; no result was computed
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("0fa380f44af066ff")
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, TransLayer: 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
from timm.models.layers import to_2tuple

torch.manual_seed(0)

class ConvBNReLU(nn.Sequential):
    def __init__(self, in_planes, out_planes, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=False, act_name='relu', is_transposed=False):
        super().__init__()
        if is_transposed:
            conv_module = nn.ConvTranspose2d
        else:
            conv_module = nn.Conv2d
        self.add_module(name='conv', module=conv_module(in_planes, out_planes, kernel_size=kernel_size, stride=to_2tuple(stride), padding=to_2tuple(padding), dilation=to_2tuple(dilation), groups=groups, bias=bias))
        self.add_module(name='bn', module=nn.BatchNorm2d(out_planes))
        if act_name is not None:
            self.add_module(name=act_name, module=_get_act_fn(act_name=act_name))

def _get_act_fn(act_name, inplace=True):
    if act_name == 'relu':
        return nn.ReLU(inplace=inplace)
    elif act_name == 'leaklyrelu':
        return nn.LeakyReLU(negative_slope=0.1, inplace=inplace)
    elif act_name == 'gelu':
        return nn.GELU()
    else:
        raise NotImplementedError

def cus_sample(x, mode='size', factors=None):
    if mode == 'size' and factors is not None:
        return torch.nn.functional.interpolate(x, size=factors, mode='nearest')
    return x

class ASPP(nn.Module):
    def __init__(self, in_dim, out_dim):
        super(ASPP, self).__init__()
        self.conv1 = ConvBNReLU(in_dim, out_dim, kernel_size=1)
        self.conv2 = ConvBNReLU(in_dim, out_dim, kernel_size=3, dilation=2, padding=2)
        self.conv3 = ConvBNReLU(in_dim, out_dim, kernel_size=3, dilation=5, padding=5)
        self.conv4 = ConvBNReLU(in_dim, out_dim, kernel_size=3, dilation=7, padding=7)
        self.conv5 = ConvBNReLU(in_dim, out_dim, kernel_size=1)
        self.fuse = ConvBNReLU(5 * out_dim, out_dim, 3, 1, 1)

    def forward(self, x):
        conv1 = self.conv1(x)
        conv2 = self.conv2(x)
        conv3 = self.conv3(x)
        conv4 = self.conv4(x)
        conv5 = self.conv5(cus_sample(x.mean((2, 3), keepdim=True), mode='size', factors=x.size()[2:]))
        return self.fuse(torch.cat((conv1, conv2, conv3, conv4, conv5), 1))

class TransLayer(nn.Module):
    def __init__(self, out_c, last_module=ASPP):
        super().__init__()
        self.c5_down = nn.Sequential(last_module(in_dim=2048, out_dim=out_c))
        self.c4_down = nn.Sequential(ConvBNReLU(1024, out_c, 3, 1, 1))
        self.c3_down = nn.Sequential(ConvBNReLU(512, out_c, 3, 1, 1))
        self.c2_down = nn.Sequential(ConvBNReLU(256, out_c, 3, 1, 1))
        self.c1_down = nn.Sequential(ConvBNReLU(64, out_c, 3, 1, 1))

    def forward(self, xs):
        assert isinstance(xs, (tuple, list))
        assert len(xs) == 5
        c1, c2, c3, c4, c5 = xs
        c5 = self.c5_down(c5)
        c4 = self.c4_down(c4)
        c3 = self.c3_down(c3)
        c2 = self.c2_down(c2)
        c1 = self.c1_down(c1)
        return (c5, c4, c3, c2, c1)

ARGS = {
    'out_c': 4,
    'last_module': ASPP
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type TransLayer — 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 TransLayer 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_fn Ran
ran; our contract draft was wrong, not the code
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("22f8cb77f8cc8af2")
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 ReLU. 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)

# The function _get_act_fn takes two parameters:
# 1. act_name: a string indicating which activation function to return
# 2. inplace: a boolean flag (default True)

# We'll test with 'relu' as a valid activation name
ARGS = {
    'act_name': 'relu',
    'inplace': 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 ReLU — a ReLU output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ReLU 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
cal_ual Ran
ran; our contract draft was wrong, not the code
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("1c4040dd702dbed1")
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 Tensor, 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create small tensors for seg_logits and seg_gts
# The function expects them to have the same shape
# It applies sigmoid, performs operations, and returns a scalar loss
batch_size = 2
spatial_dims = 4

seg_logits = torch.randn(batch_size, spatial_dims, dtype=torch.float32)
seg_gts = torch.randn(batch_size, spatial_dims, dtype=torch.float32)

ARGS = {
    "seg_logits": seg_logits,
    "seg_gts": seg_gts,
}
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 Tensor, 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
formatter_for_dict Ran
ran; our contract draft was wrong, not the code
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("a1699c744bd715fd")
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 str. 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

# Construct valid arguments for formatter_for_dict function
info_dict = {
    'model_name': 'ResNet50',
    'learning_rate': 0.001,
    'batch_size': 32,
    'epochs': 100,
    'accuracy': 0.95
}

headers = ['Parameter', 'Value']

name_length = 20
obj_str_length = 60

ARGS = {
    'info_dict': info_dict,
    'headers': headers,
    'name_length': name_length,
    'obj_str_length': obj_str_length
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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_coef Ran
ran, honoured the contract we drafted
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("1ed4fe7e2e6020fa")
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 float, 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

ARGS = {
    "iter_percentage": 0.5,
    "method": "linear"
}
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 float, 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
item2str Ran
ran; our contract draft was wrong, not the code
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("ce165bae8262bf4c")
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 str. 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 numpy as np

# Construct valid arguments for item2str function
ARGS = {
    'string': 'hello',
    'max_length': 10,
    'padding_char': ' ',
    'mode': 'left'
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
ZoomNet Not yet run
unverified (harvested, no recorded run)
lartpang/ZoomNet/methods/zoomnet/zoomnet.py
code served (permissive licence) · get_code("6e01acf2e56ffb95")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 2 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.

FunctionAlso present inEarliest-datedOthers named
cal_ual
brick of this paper's assembly: loss
1 other paper
1 later
2506.21866 (2025-06)
Dual-Perspective United Transformer for Object Segmentation in Optical Remote Sensing Images
get_coef
brick of this paper's assembly: helper
1 other paper
1 later
2506.21866 (2025-06)
Dual-Perspective United Transformer for Object Segmentation in Optical Remote Sensing Images

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

The recently proposed camouflaged object detection (COD) attempts to segment objects that are visually blended into their surroundings, which is extremely complex and difficult in real-world scenarios. Apart from high intrinsic similarity between the camouflaged objects and their background, the objects are usually diverse in scale, fuzzy in appearance, and even severely occluded. To deal with these problems, we propose a mixed-scale triplet network, Zoom-Net, which mimics the behavior of humans when observing vague images, i.e., zooming in and out. Specifically, our ZoomNet employs the zoom strategy to learn the discriminative mixed-scale semantics by the designed scale integration unit and hierarchical mixed-scale unit, which fully explores imperceptible clues between the candidate objects and background surroundings. Moreover, considering the uncertainty and ambiguity derived from indistinguishable textures, we construct a simple yet effective regularization constraint, uncertainty-aware loss, to promote the model to accurately produce predictions with higher confidence in candidate regions. Without bells and whistles, our proposed highly task-friendly model consistently surpasses the existing 23 state-of-the-art methods on four public datasets. Besides, the superior performance over the recent cuttingedge models on the SOD task also verifies the effectiveness and generality of our model. The code will be available at https://github.com/lartpang/ZoomNet.

For agents

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

get_harvested_code_for_paper("2203.02688")
get_code_for_paper("2203.02688")
have("2203.02688")

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

curl https://syntology.ai/api/ran/2203.02688.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.02688.svg)](https://syntology.ai/paper/2203.02688)

Connect an agent — have() is free.