SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.08478 · ICCV · 2021

Learning Anchored Unsigned Distance Functions with Gradient Direction Alignment for Single-view Garment Reconstruction

Ling Shao, Shengcai Liao, Wenhao Wang, Fang Zhao

arXiv · PDF · Open in the Atlas

Code that ran

10 of 18 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; 0 honoured the contract we drafted, 0 violated it and 6 have no contract verdict; 4 failures of our instrument; 8 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/2108.08478.json · badge: /api/ran/2108.08478.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
zhaofang0627/anchorudf canonical (classifier label)
classifier basis: census
10 ran · 8 no recorded run
FunctionStatusWhere it lives
adjust_learning_rate Ran
ran on a synthesized input
zhaofang0627/anchorudf/lib/net_util.py
code served (permissive licence) · get_code("c909dbba49acbe18")
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 torch
import torch.nn as nn
from torch.optim import SGD

torch.manual_seed(0)

# Create a simple model
model = nn.Linear(4, 2)

# Create an optimizer
optimizer = SGD(model.parameters(), lr=0.1)

# Define arguments for adjust_learning_rate
ARGS = {
    'optimizer': optimizer,
    'epoch': 5,
    'lr': 0.1,
    'schedule': [5, 10, 15],
    'gamma': 0.1
}
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 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
batch_eval Ran
ran on a synthesized input
zhaofang0627/anchorudf/lib/sdf.py
code served (permissive licence) · get_code("1c425b0d9e08d0c3")
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 ndarray, ndim 1, dtype family float, shape (1000,) 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

# Define a simple eval function that takes points and returns SDF values
def simple_eval_func(points):
    """
    Takes points of shape (3, num_pts) and returns SDF values of shape (num_pts,)
    """
    # Simple mock SDF: just sum the coordinates
    return np.sum(points, axis=0)

# Create sample points
# Shape: (3, 1000) - 3D points, 1000 of them
points = np.random.randn(3, 1000).astype(np.float32)

ARGS = {
    "points": points,
    "eval_func": simple_eval_func,
    "num_samples": 512
}
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 ndarray, ndim 1, dtype family float, shape (1000,) — shape (1000,) 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
create_grid Ran
ran on a synthesized input
zhaofang0627/anchorudf/lib/sdf.py
code served (permissive licence) · get_code("190c784e12e69338")
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

# Set random seed for reproducibility
np.random.seed(0)

# Define arguments for create_grid function
ARGS = {
    "resX": 4,
    "resY": 5,
    "resZ": 6,
    "b_min": np.array([0.0, 0.0, 0.0]),
    "b_max": np.array([1.0, 1.0, 1.0]),
    "transform": np.array([
        [1.0, 0.0, 0.0, 0.1],
        [0.0, 1.0, 0.0, 0.2],
        [0.0, 0.0, 1.0, 0.3],
        [0.0, 0.0, 0.0, 1.0]
    ])
}
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
create_grid_points_from_bounds Ran
ran on a synthesized input
zhaofang0627/anchorudf/lib/mesh_util.py
code served (permissive licence) · get_code("30b375d99e7c6628")
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 ndarray, ndim 2, dtype family float, shape (27, 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 numpy as np

ARGS = {
    "minimun": 0.0,
    "maximum": 1.0,
    "res": 3
}
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 ndarray, ndim 2, dtype family float, shape (27, 3) — shape (27, 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
divergence Ran
ran; our contract draft was wrong, not the code
zhaofang0627/anchorudf/lib/diff_operators.py
code served (permissive licence) · get_code("87e822f4bcfbdac7")
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 2, dtype family float, shape (2, 1) 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)

# The divergence function computes the divergence of a vector field y with respect to x
# It requires:
# - y: a tensor with shape [..., n] where n is the output dimension
# - x: a tensor that y depends on, requires grad

# Create a small input tensor x with requires_grad=True
x = torch.randn(2, 4, requires_grad=True)

# Create y as a function of x
# y should have shape [..., output_dim]
# We'll create y with output dimension 3 (distinct from input dim 4)
# y = some function of x, e.g., a simple linear-like transformation
y = torch.sum(x, dim=1, keepdim=True) * torch.ones(2, 3)
y.requires_grad_(True)

# Alternatively, if y needs to be explicitly computed from x:
# Let's make y depend on x through autograd graph
y = x[:, :3] * 2.0 + x[:, 1:4].sum(dim=1, keepdim=True)

ARGS = {
    "y": y,
    "x": x,
}
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 2, dtype family float, shape (2, 1) — shape (2, 1) 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
eval_grid Ran
ran on a synthesized input
zhaofang0627/anchorudf/lib/sdf.py
code served (permissive licence) · get_code("2e4f8c7dfe5817f0")
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 ndarray, ndim 3, dtype family float, shape (4, 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 numpy as np

# Define a simple eval function that takes a point cloud and returns SDF values
def simple_eval_func(points_batch):
    """
    Takes points of shape (3, N) and returns N SDF values.
    For testing, return a simple function based on distance from origin.
    """
    # points_batch shape: (3, num_points)
    distances = np.sqrt(np.sum(points_batch ** 2, axis=0))
    return distances - 1.0  # Simple SDF: distance to unit sphere

# Create coordinate grid
# Shape should be (3, res1, res2, res3) for eval_grid
resolution = (4, 5, 6)  # Small distinct dimensions
coords = np.random.randn(3, *resolution).astype(np.float32)

ARGS = {
    "coords": coords,
    "eval_func": simple_eval_func,
    "num_samples": 512 * 512 * 512
}
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 ndarray, ndim 3, dtype family float, shape (4, 5, 6) — shape (4, 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
get_verts_rgb_colors Ran
ran on a synthesized input
zhaofang0627/anchorudf/lib/colab_util.py
code served (permissive licence) · get_code("add73ec2e2c2eccb")
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 ndarray, ndim 3, dtype family float, shape (1, 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 numpy as np
import tempfile
import os

# Create a temporary OBJ file with vertex data including RGB colors
temp_dir = tempfile.mkdtemp()
obj_path = os.path.join(temp_dir, "test_model.obj")

# Write a simple OBJ file with 7-element lines (vertex data with RGB)
with open(obj_path, 'w') as f:
    f.write("v 0.0 0.0 0.0 1.0 0.0 0.0\n")  # 7 elements: x y z r g b
    f.write("v 1.0 1.0 1.0 0.0 1.0 0.0\n")
    f.write("v 2.0 2.0 2.0 0.0 0.0 1.0\n")
    f.write("# this is a comment line\n")
    f.write("v 3.0 3.0 3.0 1.0 1.0 0.0\n")

ARGS = {
    "obj_path": obj_path
}
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 ndarray, ndim 3, dtype family float, shape (1, 4, 3) — shape (1, 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
hessian Ran
ran; our contract draft was wrong, not the code
zhaofang0627/anchorudf/lib/diff_operators.py
code served (permissive licence) · get_code("fd089b9f4b8749ab")
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
torch.manual_seed(0)

# The function expects:
# y: shape (meta_batch_size, num_observations, channels)
# x: shape (meta_batch_size, num_observations, 2)
# Both need to require grad for autograd to work

meta_batch_size = 2
num_observations = 3
channels = 4
x_dim = 2

# x needs to be differentiable
x = torch.randn(meta_batch_size, num_observations, x_dim, requires_grad=True, dtype=torch.float32)

# y should be a function of x to compute meaningful gradients
# Create a simple function: y = x^2 summed in a way that gives us channels outputs
y = torch.sum(x.unsqueeze(-1) ** 2, dim=2)  # shape (meta_batch_size, num_observations, 1)
y = y.expand(meta_batch_size, num_observations, channels)  # expand to (meta_batch_size, num_observations, channels)
y.requires_grad_(True)

ARGS = {
    "y": y,
    "x": x,
}
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 — 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
index Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zhaofang0627/anchorudf/lib/geometry.py
code served (permissive licence) · get_code("9e75a53ec16ff31a")
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 3, dtype family float, shape (2, 4, 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 torch.nn.functional as F

torch.manual_seed(0)

# feat: [B, C, H, W] image features
# B=2 (batch), C=4 (channels), H=3 (height), W=5 (width)
feat = torch.randn(2, 4, 3, 5)

# uv: [B, 2, N] uv coordinates in the image plane, range [-1, 1]
# B=2 (batch), 2 (u and v coordinates), N=6 (number of points)
uv = torch.rand(2, 2, 6) * 2 - 1  # range [-1, 1]

ARGS = {
    "feat": feat,
    "uv": uv
}
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 3, dtype family float, shape (2, 4, 6) — shape (2, 4, 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
laplace Ran
ran; our contract draft was wrong, not the code
zhaofang0627/anchorudf/lib/diff_operators.py
code served (permissive licence) · get_code("a2be4be6558869af")
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 2, dtype family float, shape (2, 1) 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 a small tensor y that depends on x via autograd graph
# y needs to have a last dimension so we can iterate over it in divergence()
# Shape: (batch=2, feature_dim=4)
x = torch.randn(2, 4, dtype=torch.float32, requires_grad=True)

# Create y as a function of x so the autograd graph is built
# y should also have shape compatible with the divergence function
# We'll make y = x^2 to create a simple dependency
y = (x ** 2).sum(dim=1, keepdim=True)  # Shape: (2, 1)

ARGS = {
    "y": y,
    "x": x,
}
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 2, dtype family float, shape (2, 1) — shape (2, 1) 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
orthogonal Not yet run
unverified (harvested, no recorded run)
zhaofang0627/anchorudf/lib/geometry.py
code served (permissive licence) · get_code("c923c46cac70f2c7")
How it ran

no invocation recorded.

perspective Not yet run
unverified (harvested, no recorded run)
zhaofang0627/anchorudf/lib/geometry.py
code served (permissive licence) · get_code("77c9954623b479a9")
How it ran

no invocation recorded.

reconstruction Not yet run
unverified (harvested, no recorded run)
zhaofang0627/anchorudf/lib/mesh_util.py
code served (permissive licence) · get_code("fa0763997c45084b")
How it ran

no invocation recorded.

reconstruction_anchor Not yet run
unverified (harvested, no recorded run)
zhaofang0627/anchorudf/lib/mesh_util.py
code served (permissive licence) · get_code("6105ec53e0ac47ed")
How it ran

no invocation recorded.

reshape_multiview_tensors Not yet run
unverified (harvested, no recorded run)
zhaofang0627/anchorudf/lib/net_util.py
code served (permissive licence) · get_code("4cf11c5a14b1f039")
How it ran

no invocation recorded.

reshape_sample_tensor Not yet run
unverified (harvested, no recorded run)
zhaofang0627/anchorudf/lib/net_util.py
code served (permissive licence) · get_code("79509ef34e0af8e4")
How it ran

no invocation recorded.

save_samples_rgb Not yet run
unverified (harvested, no recorded run)
zhaofang0627/anchorudf/lib/sample_util.py
code served (permissive licence) · get_code("757063dc19881fa9")
How it ran

no invocation recorded.

save_samples_truncted_prob Not yet run
unverified (harvested, no recorded run)
zhaofang0627/anchorudf/lib/sample_util.py
code served (permissive licence) · get_code("d9c92c86878b41e7")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 6 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 5 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
laplace 6 other papers
2 dated earlier, 4 later
2006.09661 (2020-06)
Implicit Neural Representations with Periodic Activation Functions
2007.09601 (2020-07)
Hypersolvers: Toward Fast Continuous-Depth Models
2303.14092 (2023-03)
NeuFace: Realistic 3D Neural Face Rendering from Multi-view Images
and 3 more not named here
divergence 4 other papers
1 dated earlier, 3 later
2006.09661 (2020-06)
Implicit Neural Representations with Periodic Activation Functions
2303.14092 (2023-03)
NeuFace: Realistic 3D Neural Face Rendering from Multi-view Images
2312.04028 (2023-12)
ImFace++: A Sophisticated Nonlinear 3D Morphable Face Model with Implicit Neural Representations
and 1 more not named here
hessian 3 other papers
1 dated earlier, 2 later
2006.09661 (2020-06)
Implicit Neural Representations with Periodic Activation Functions
2303.14092 (2023-03)
NeuFace: Realistic 3D Neural Face Rendering from Multi-view Images
2312.04028 (2023-12)
ImFace++: A Sophisticated Nonlinear 3D Morphable Face Model with Implicit Neural Representations
reshape_sample_tensor 2 other papers
2 later
2309.13524 (2023-09)
Global-correlated 3D-decoupling Transformer for Clothed Avatar Reconstruction
2312.06704 (2023-12)
SIFU: Side-view Conditioned Implicit Function for Real-world Usable Clothed Human Reconstruction
adjust_learning_rate 1 other paper
1 dated earlier
1912.08265 (2019-12)
Learning from Synthetic Animals
index 1 other paper
1 later
2204.02445 (2022-04)
CHORE: Contact, Human and Object REconstruction from a single RGB image

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

While single-view 3D reconstruction has made significant progress benefiting from deep shape representations in recent years, garment reconstruction is still not solved well due to open surfaces, diverse topologies and complex geometric details. In this paper, we propose a novel learnable Anchored Unsigned Distance Function (AnchorUDF) representation for 3D garment reconstruction from a single image. AnchorUDF represents 3D shapes by predicting unsigned distance fields (UDFs) to enable open garment surface modeling at arbitrary resolution. To capture diverse garment topologies, AnchorUDF not only computes pixel-aligned local image features of query points, but also leverages a set of anchor points located around the surface to enrich 3D position features for query points, which provides stronger 3D space context for the distance function. Furthermore, in order to obtain more accurate point projection direction at inference, we explicitly align the spatial gradient direction of AnchorUDF with the ground-truth direction to the surface during training. Extensive experiments on two public 3D garment datasets, i.e., MGN and Deep Fashion3D, demonstrate that AnchorUDF achieves the state-of-the-art performance on single-view garment reconstruction. Code is available at https://github. com/zhaofang0627/AnchorUDF.

For agents

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

get_harvested_code_for_paper("2108.08478")
get_code_for_paper("2108.08478")
have("2108.08478")

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

curl https://syntology.ai/api/ran/2108.08478.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.08478.svg)](https://syntology.ai/paper/2108.08478)

Connect an agent — have() is free.