SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.08017 · ICCV · 2021

Deep Hybrid Self-Prior for Full 3D Mesh Generation

Yanwei Fu, Zhaopeng Cui, Yinda Zhang, Xingkui Wei, Zhengqing Chen

arXiv · PDF · Open in the Atlas

Code that ran

4 of 6 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (3 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 1 failure of our instrument; 2 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/2108.08017.json · badge: /api/ran/2108.08017.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
weixk2015/dhsp3d community
Papers with Code archive
4 ran · 2 no recorded run
FunctionStatusWhere it lives
MeshConv Ran
ran on a synthesized input · constructed an object; no result was computed
weixk2015/dhsp3d/models/networks.py
code served (permissive licence) · get_code("cfa2853745af94de")
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, MeshConv: 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)

# Create a minimal mesh data structure that has the required attributes
class MockMesh:
    def __init__(self, edges_count=8, gemm_edges=None):
        self.edges_count = edges_count
        if gemm_edges is None:
            # gemm_edges should be edges_count x 4 (4 one-ring neighbors per edge)
            self.gemm_edges = torch.randint(0, edges_count, (edges_count, 4)).tolist()
        else:
            self.gemm_edges = gemm_edges

# Create mock mesh with 8 edges
mock_mesh = MockMesh(edges_count=8)

# Create edge features: (Batch=2, Features=4, Edges=8)
# The forward pass will squeeze the last dimension, so we add one more
edge_features = torch.randn(2, 4, 8, 1)

# Create a list of mesh objects (one per batch item)
mesh_list = [mock_mesh, mock_mesh]

ARGS = {
    "in_channels": 4,
    "out_channels": 3,
    "k": 5,
    "bias": 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 MeshConv — 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 MeshConv 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
MeshPool Ran
ran on a synthesized input · constructed an object; no result was computed
weixk2015/dhsp3d/models/networks.py
code served (permissive licence) · get_code("aa6b6ff3feace7bc")
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, MeshPool: 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 numpy as np

torch.manual_seed(0)

class SimpleMesh:
    def __init__(self, edges_count=8):
        self.edges_count = edges_count
        self.gemm_edges = [[-1, -1, -1, -1] for _ in range(edges_count)]
        
    def clean(self, mask, edge_groups):
        pass
    
    def merge_vertices(self, edge_id):
        pass

# Create small test data
fe = torch.randn(2, 4, 8, dtype=torch.float32)  # batch=2, features=4, edges=8
mesh1 = SimpleMesh(edges_count=8)
mesh2 = SimpleMesh(edges_count=8)
meshes = [mesh1, mesh2]

target = 4
multi_thread = False

ARGS = {
    'target': target,
    'multi_thread': multi_thread,
}
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 MeshPool — 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 MeshPool 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
MeshUnion Ran
ran on a synthesized input · constructed an object; no result was computed
weixk2015/dhsp3d/models/networks.py
code served (permissive licence) · get_code("855f0326199df23c")
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, MeshUnion: 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 numpy as np
from torch.nn import ConstantPad2d

torch.manual_seed(0)

ARGS = {
    "n": 5,
    "device": torch.device('cpu')
}
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 MeshUnion — 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 MeshUnion 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_norm_args Ran
ran; our contract draft was wrong, not the code
weixk2015/dhsp3d/models/networks.py
code served (permissive licence) · get_code("a5fee593f2be29f3")
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 list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from functools import partial

torch.manual_seed(0)

# Define a NoNorm class for testing
class NoNorm(nn.Module):
    def __init__(self, fake=True):
        super().__init__()
        self.fake = fake

# Test case 1: NoNorm
norm_layer_1 = NoNorm
nfeats_list_1 = [16, 32, 64]

# Test case 2: GroupNorm via functools.partial
norm_layer_2 = partial(nn.GroupNorm)
nfeats_list_2 = [8, 16, 32]

# Test case 3: BatchNorm via functools.partial
norm_layer_3 = partial(nn.BatchNorm2d)
nfeats_list_3 = [4, 8, 16]

# Use the first test case for ARGS
ARGS = {
    'norm_layer': norm_layer_1,
    'nfeats_list': nfeats_list_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 list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
MResConv Not yet run
unverified (harvested, no recorded run)
weixk2015/dhsp3d/models/networks.py
code served (permissive licence) · get_code("10ff731d52c8169a")
How it ran

no invocation recorded.

MeshConvNet Not yet run
unverified (harvested, no recorded run)
weixk2015/dhsp3d/models/networks.py
code served (permissive licence) · get_code("2ee783352da9a68c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 6 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Ours Figure 1. We propose to utilize the deep hybrid 2D-3D self-prior in neural networks to generate the high-quality textured 3D mesh model from the sparse colored point cloud.

For agents

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

get_harvested_code_for_paper("2108.08017")
get_code_for_paper("2108.08017")
have("2108.08017")

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

curl https://syntology.ai/api/ran/2108.08017.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.08017.svg)](https://syntology.ai/paper/2108.08017)

Connect an agent — have() is free.