SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2204.07346 · ECCV · 2022

MVSTER: Epipolar Transformer for Efficient Multi-View Stereo

Zheng Zhu, Xiaofeng Wang, Xu Chi, Guan Huang, Xingang Wang, Yun Ye, Fangbo Qin, Yijia He

arXiv · PDF · Open in the Atlas

Code that ran

10 of 17 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (5 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 5 failures of our instrument; 7 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/2204.07346.json · badge: /api/ran/2204.07346.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
JeffWang987/MVSTER community
Papers with Code archive
10 ran · 7 no recorded run
FunctionStatusWhere it lives
Conv2d Ran
ran on a synthesized input · constructed an object; no result was computed
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("3aafeb0680c73abf")
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, Conv2d: 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 = {
    'in_channels': 3,
    'out_channels': 16,
    'kernel_size': 3,
    'stride': 1,
    'relu': True,
    'bn_momentum': 0.1,
    'init_method': 'xavier',
    'gn': False,
    'group_channel': 8,
}
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 Conv2d — 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 Conv2d 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
ConvBnReLU3D Ran
ran on a synthesized input · constructed an object; no result was computed
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("2a8c0668448c2f85")
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, ConvBnReLU3D: 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)

# ConvBnReLU3D parameters:
# - in_channels: int
# - out_channels: int
# - kernel_size: int (default 3)
# - stride: int (default 1)
# - pad: int (default 1)

ARGS = {
    "in_channels": 3,
    "out_channels": 16,
    "kernel_size": 3,
    "stride": 1,
    "pad": 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 ConvBnReLU3D — 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 ConvBnReLU3D 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
FPN4 Ran
ran on a synthesized input · constructed an object; no result was computed
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("38d084809ff8b324")
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, FPN4: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Define stub classes for missing dependencies
class DeformConvPack(nn.Module):
    def __init__(self, in_channels, out_channels, kernel_size=3, stride=1, padding=1, groups=1, deformable_groups=1, bias=False, im2col_step=16):
        super(DeformConvPack, self).__init__()
        self.conv = nn.Conv2d(in_channels, out_channels, kernel_size, stride=stride, padding=padding, groups=groups, bias=bias)
    
    def forward(self, x):
        return self.conv(x)

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

ARGS = {
    'base_channels': 8,
    'gn': False,
    'dcn': 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 FPN4 — 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 FPN4 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
homo_warping Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("0335666ff3549d14")
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 5, dtype family float, shape (2, 4, 3, 5, 6) 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)

# Small tensor dimensions: B=2, C=4, num_depth=3, Hr=5, Wr=6, Hs=5, Ws=6
B = 2
C = 4
num_depth = 3
Hr = 5
Wr = 6
Hs = 5
Ws = 6

# src_fea: shape (B, C, Hs, Ws) for 4D case
src_fea = torch.randn(B, C, Hs, Ws)

# src_proj: shape (B, 4, 4) - source camera projection matrix
src_proj = torch.randn(B, 4, 4)

# ref_proj: shape (B, 4, 4) - reference camera projection matrix
ref_proj = torch.randn(B, 4, 4)
# Make it invertible
ref_proj = ref_proj + 5 * torch.eye(4).unsqueeze(0)

# depth_values: shape (B, num_depth, Hr, Wr)
depth_values = torch.randn(B, num_depth, Hr, Wr).abs() + 0.5

# vis_ETA: boolean flag for visualization
vis_ETA = False

# fn: filename for saving (not used when vis_ETA=False)
fn = None

ARGS = {
    'src_fea': src_fea,
    'src_proj': src_proj,
    'ref_proj': ref_proj,
    'depth_values': depth_values,
    'vis_ETA': vis_ETA,
    'fn': fn
}
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 Tensor, ndim 5, dtype family float, shape (2, 4, 3, 5, 6) — shape (2, 4, 3, 5, 6) 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
init_inverse_range Ran
ran; our contract draft was wrong, not the code
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("9dc0b1027a30eb55")
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 4, dtype family float, shape (2, 4, 3, 5) 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

torch.manual_seed(0)

# Create small tensors with distinct sizes for different dimensions
batch_size = 2
num_depth_samples = 4
height = 3
width = 5

# cur_depth: shape (batch_size, num_depth_samples)
# This tensor holds depth values at different sample points
cur_depth = torch.tensor([
    [0.5, 1.0, 1.5, 2.0],
    [0.6, 1.2, 1.8, 2.4]
], dtype=torch.float32)

# Number of depth hypothesis points to generate
ndepths = 4

# Device and dtype
device = torch.device('cpu')
dtype = torch.float32

# Height and Width
H = 3
W = 5

ARGS = {
    'cur_depth': cur_depth,
    'ndepths': ndepths,
    'device': device,
    'dtype': dtype,
    'H': H,
    'W': W
}
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 4, dtype family float, shape (2, 4, 3, 5) — shape (2, 4, 3, 5) 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
init_range Ran
ran; our contract draft was wrong, not the code
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("c72a4a2293997c31")
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 4, dtype family float, shape (2, 4, 3, 2) 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

torch.manual_seed(0)

# Construct arguments for init_range function
# Parameters: cur_depth, ndepths, device, dtype, H, W

# cur_depth: tensor with shape (batch, 2) - min and max depth values per batch
# The function accesses cur_depth[:, 0] and cur_depth[:, -1]
cur_depth = torch.tensor([
    [0.5, 5.0],
    [1.0, 8.0]
], dtype=torch.float32)

# ndepths: number of depth samples
ndepths = 4

# device: cpu device
device = 'cpu'

# dtype: float32
dtype = torch.float32

# H, W: height and width of the output depth map
H = 3
W = 2

ARGS = {
    'cur_depth': cur_depth,
    'ndepths': ndepths,
    'device': device,
    'dtype': dtype,
    'H': H,
    'W': W
}
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 4, dtype family float, shape (2, 4, 3, 2) — shape (2, 4, 3, 2) 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
reg2d Ran
ran on a synthesized input · constructed an object; no result was computed
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("d6051468a8c98437")
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, reg2d: 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
import sys
from unittest.mock import MagicMock

torch.manual_seed(0)

# Create mock ConvBnReLU3D class that the module import will use
class ConvBnReLU3D(nn.Module):
    def __init__(self, in_channels, out_channels, kernel_size=(3, 3, 3), stride=(1, 1, 1), pad=(1, 1, 1)):
        super(ConvBnReLU3D, self).__init__()
        self.conv = nn.Conv3d(in_channels, out_channels, kernel_size=kernel_size, stride=stride, padding=pad, bias=False)
        self.bn = nn.BatchNorm3d(out_channels)
        self.relu = nn.ReLU(inplace=True)
    
    def forward(self, x):
        return self.relu(self.bn(self.conv(x)))

# Create a mock module that has ConvBnReLU3D
mock_module = MagicMock()
mock_module.ConvBnReLU3D = ConvBnReLU3D
mock_module.ConvBnReLU3D3D = ConvBnReLU3D

# Patch importlib to return our mock module
import importlib
original_import_module = importlib.import_module

def mock_import_module(name):
    if name == 'models.mvs4net_utils':
        return mock_module
    return original_import_module(name)

importlib.import_module = mock_import_module

# Create a small test input tensor
# Shape: (batch, channels, depth, height, width)
# Using distinct sizes: batch=2, channels=128, depth=4, height=8, width=16
test_input = torch.randn(2, 128, 4, 8, 16)

ARGS = {
    'input_channel': 128,
    'base_channel': 32,
    'conv_name': 'ConvBnReLU3D'
}
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 reg2d — 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 reg2d 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
reg3d Ran
ran on a synthesized input · constructed an object; no result was computed
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("23a0bbc6e0990065")
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, reg3d: 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
import torch.nn.functional as F

torch.manual_seed(0)

class ConvBnReLU3D(nn.Module):

    def __init__(self, in_channels, out_channels, kernel_size=3, stride=1, pad=1):
        super(ConvBnReLU3D, self).__init__()
        self.conv = nn.Conv3d(in_channels, out_channels, kernel_size, stride=stride, padding=pad, bias=False)
        self.bn = nn.BatchNorm3d(out_channels)

    def forward(self, x):
        return F.relu(self.bn(self.conv(x)), inplace=True)

class reg3d(nn.Module):

    def __init__(self, in_channels, base_channels, down_size=3):
        super(reg3d, self).__init__()
        self.down_size = down_size
        self.conv0 = ConvBnReLU3D(in_channels, base_channels, kernel_size=3, pad=1)
        self.conv1 = ConvBnReLU3D(base_channels, base_channels * 2, kernel_size=3, stride=2, pad=1)
        self.conv2 = ConvBnReLU3D(base_channels * 2, base_channels * 2)
        if down_size >= 2:
            self.conv3 = ConvBnReLU3D(base_channels * 2, base_channels * 4, kernel_size=3, stride=2, pad=1)
            self.conv4 = ConvBnReLU3D(base_channels * 4, base_channels * 4)
        if down_size >= 3:
            self.conv5 = ConvBnReLU3D(base_channels * 4, base_channels * 8, kernel_size=3, stride=2, pad=1)
            self.conv6 = ConvBnReLU3D(base_channels * 8, base_channels * 8)
            self.conv7 = nn.Sequential(nn.ConvTranspose3d(base_channels * 8, base_channels * 4, kernel_size=3, padding=1, output_padding=1, stride=2, bias=False), nn.BatchNorm3d(base_channels * 4), nn.ReLU(inplace=True))
        if down_size >= 2:
            self.conv9 = nn.Sequential(nn.ConvTranspose3d(base_channels * 4, base_channels * 2, kernel_size=3, padding=1, output_padding=1, stride=2, bias=False), nn.BatchNorm3d(base_channels * 2), nn.ReLU(inplace=True))
        self.conv11 = nn.Sequential(nn.ConvTranspose3d(base_channels * 2, base_channels, kernel_size=3, padding=1, output_padding=1, stride=2, bias=False), nn.BatchNorm3d(base_channels), nn.ReLU(inplace=True))
        self.prob = nn.Conv3d(base_channels, 1, 3, stride=1, padding=1, bias=False)

    def forward(self, x):
        if self.down_size == 3:
            conv0 = self.conv0(x)
            conv2 = self.conv2(self.conv1(conv0))
            conv4 = self.conv4(self.conv3(conv2))
            x = self.conv6(self.conv5(conv4))
            x = conv4 + self.conv7(x)
            x = conv2 + self.conv9(x)
            x = conv0 + self.conv11(x)
            x = self.prob(x)
        elif self.down_size == 2:
            conv0 = self.conv0(x)
            conv2 = self.conv2(self.conv1(conv0))
            x = self.conv4(self.conv3(conv2))
            x = conv2 + self.conv9(x)
            x = conv0 + self.conv11(x)
            x = self.prob(x)
        else:
            conv0 = self.conv0(x)
            x = self.conv2(self.conv1(conv0))
            x = conv0 + self.conv11(x)
            x = self.prob(x)
        return x.squeeze(1)

model = reg3d(in_channels=3, base_channels=8, down_size=3)
model.eval()

input_tensor = torch.randn(2, 3, 4, 4, 4)

ARGS = {
    "in_channels": 3,
    "base_channels": 8,
    "down_size": 3
}
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 reg3d — 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 reg3d 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
schedule_inverse_range Ran
ran; our contract draft was wrong, not the code
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("afbfac7ff243b94d")
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 4, dtype family float, shape (2, 4, 8, 6) 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

torch.manual_seed(0)

# Analyze function usage:
# - inverse_min_depth: tensor with .device and .dtype attributes, used in indexing [:, None, :, :]
# - inverse_max_depth: tensor with .device and .dtype attributes, used in indexing [:, None, :, :]
# - ndepths: int, used in arange and reshape
# - H: int, used in repeat and interpolate
# - W: int, used in repeat and interpolate

batch_size = 2
height = 8
width = 6
num_depths = 4
spatial_h = height // 2  # 4
spatial_w = width // 2   # 3

# Create inverse_min_depth and inverse_max_depth tensors
# They need shape (batch, H//2, W//2) based on how they're used in the function
inverse_min_depth = torch.ones(batch_size, spatial_h, spatial_w, dtype=torch.float32)
inverse_max_depth = torch.ones(batch_size, spatial_h, spatial_w, dtype=torch.float32) * 0.5

ARGS = {
    "inverse_min_depth": inverse_min_depth,
    "inverse_max_depth": inverse_max_depth,
    "ndepths": num_depths,
    "H": height,
    "W": width,
}
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 4, dtype family float, shape (2, 4, 8, 6) — shape (2, 4, 8, 6) 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
schedule_range Ran
ran; our contract draft was wrong, not the code
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("1221de2433902805")
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 4, dtype family float, shape (2, 5, 4, 3) 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.functional as F

torch.manual_seed(0)

# Analyze the function signature and body:
# - cur_depth: tensor used in arithmetic operations, needs shape (B, H, W)
# - ndepth: integer, number of depth samples
# - depth_inteval_pixel: tensor with shape (B, 1, 1) based on indexing [:, None, None]
# - H: integer, height dimension
# - W: integer, width dimension

# Create small tensors with distinct dimensions
batch_size = 2
height = 4
width = 3
num_depths = 5

# cur_depth: (batch, height, width) - depth values at each spatial location
cur_depth = torch.randn(batch_size, height, width, dtype=torch.float32, requires_grad=False)

# ndepth: number of depth samples to generate
ndepth = num_depths

# depth_inteval_pixel: (batch,) - interval per batch, will be indexed with [:, None, None]
depth_inteval_pixel = torch.ones(batch_size, dtype=torch.float32) * 0.1

# H and W dimensions
H = height
W = width

ARGS = {
    'cur_depth': cur_depth,
    'ndepth': ndepth,
    'depth_inteval_pixel': depth_inteval_pixel,
    'H': H,
    'W': W,
}
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 4, dtype family float, shape (2, 5, 4, 3) — shape (2, 5, 4, 3) 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
ASFF Not yet run
unverified (harvested, no recorded run)
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("71a82dc67c29d4b3")
How it ran

no invocation recorded.

MVS4net Not yet run
unverified (harvested, no recorded run)
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("594ae94979f60345")
How it ran

no invocation recorded.

NA_DCN Not yet run
unverified (harvested, no recorded run)
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("20530c2983c968db")
How it ran

no invocation recorded.

init_bn Not yet run
unverified (harvested, no recorded run)
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("e59939cd1cb63bd9")
How it ran

no invocation recorded.

init_uniform Not yet run
unverified (harvested, no recorded run)
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("a3982dbf26fc8521")
How it ran

no invocation recorded.

mono_depth_decoder Not yet run
unverified (harvested, no recorded run)
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("e8f23f1f97acdbff")
How it ran

no invocation recorded.

stagenet Not yet run
unverified (harvested, no recorded run)
JeffWang987/MVSTER/models/MVS4Net.py
code served (permissive licence) · get_code("66e954624f370968")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 5 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 12 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
init_inverse_range
brick of this paper's assembly: helper
2 other papers
2 later
Zhang_GeoMVSNet_Learning_Multi-View_Stereo_With_Geometry_Perception_CVPR_2023_paper (2023 (year only))
GeoMVSNet: Learning Multi-View Stereo with Geometry Perception
2309.17218 (2023-09)
When Epipolar Constraint Meets Non-local Operators in Multi-View Stereo
init_range
brick of this paper's assembly: helper
2 other papers
2 later
2208.02541 (2022-08)
MVSFormer: Multi-View Stereo by Learning Robust Image Features and Temperature-based Depth
2309.17218 (2023-09)
When Epipolar Constraint Meets Non-local Operators in Multi-View Stereo
init_bn
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
2112.05999 (2021-12)
Curvature-guided dynamic scale networks for Multi-view Stereo
init_uniform
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
2112.05999 (2021-12)
Curvature-guided dynamic scale networks for Multi-view Stereo
schedule_inverse_range
brick of this paper's assembly: helper
1 other paper
1 later
Zhang_GeoMVSNet_Learning_Multi-View_Stereo_With_Geometry_Perception_CVPR_2023_paper (2023 (year only))
GeoMVSNet: Learning Multi-View Stereo with Geometry Perception

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

methods warp source images into the reference camera frustum to form 3D volumes, which are fused as a cost volume to be regularized by subsequent networks. The fusing step plays a vital role in bridging 2D semantics and 3D spatial associations. However, previous methods utilize extra networks to learn 2D information as fusing cues, underusing 3D spatial correlations and bringing additional computation costs. Therefore, we present MVSTER, which leverages the proposed epipolar Transformer to learn both 2D semantics and 3D spatial associations efficiently. Specifically, the epipolar Transformer utilizes a detachable monocular depth estimator to enhance 2D semantics and uses cross-attention to construct data-dependent 3D associations along epipolar line. Additionally, MVSTER is built in a cascade structure, where entropy-regularized optimal transport is leveraged to propagate finer depth estimations in each stage. Extensive experiments show MVSTER achieves state-ofthe-art reconstruction performance with significantly higher efficiency: Compared with MVSNet and CasMVSNet, our MVSTER achieves 34% and 14% relative improvements on the DTU benchmark, with 80% and 51% relative reductions in running time. MVSTER also ranks first on Tanks&Temples-Advanced among all published works. Code is available at https://github.com/JeffWang987/MVSTER.

For agents

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

get_harvested_code_for_paper("2204.07346")
get_code_for_paper("2204.07346")
have("2204.07346")

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

curl https://syntology.ai/api/ran/2204.07346.json

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

[![Syntology run record](https://syntology.ai/api/ran/2204.07346.svg)](https://syntology.ai/paper/2204.07346)

Connect an agent — have() is free.