SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.04444 · ICCV · 2021

SnowflakeNet: Point Cloud Completion by Snowflake Point Deconvolution with Skip-Transformer

Wen Zheng, Xin Wen, Pengfei Wan, Yu-Shen Liu, Zhizhong Han, Yan-Pei Cao, Peng Xiang

arXiv · PDF · Open in the Atlas

Code that ran

8 of 16 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 6 have no contract verdict; 2 failures of our instrument; 8 no recorded run), measured 2026-11-21; 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/2108.04444.json · badge: /api/ran/2108.04444.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
allenxiangx/snowflakenet marked official in Papers with Code
Papers with Code archive
8 ran · 8 no recorded run
FunctionStatusWhere it lives
Conv2d Ran
ran on a synthesized input · constructed an object; no result was computed
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("2f63e227a64d6745")
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)

# The Conv2d class is a nn.Module that wraps nn.Conv2d with optional BatchNorm2d
# and an activation function. To test instantiation, we need to provide:
# - in_channel: number of input channels (int)
# - out_channel: number of output channels (int)
# - kernel_size: kernel size tuple (default (1, 1))
# - stride: stride tuple (default (1, 1))
# - if_bn: whether to use batch norm (default True)
# - activation_fn: activation function or None (default torch.relu)

ARGS = {
    'in_channel': 3,
    'out_channel': 16,
    'kernel_size': (3, 3),
    'stride': (1, 1),
    'if_bn': True,
    'activation_fn': torch.relu
}
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
GroupingOperation Ran
ran on a synthesized input · constructed an object; no result was computed
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("f70e6d503b13d807")
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, GroupingOperation: 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 unittest.mock import MagicMock

torch.manual_seed(0)

# Mock the _ext module since it's a C++ extension
_ext = MagicMock()

# Create mock functions that return appropriately shaped tensors
def mock_group_points(features, idx):
    # features: (B, C, N)
    # idx: (B, npoint, nsample)
    # output: (B, C, npoint, nsample)
    B, C, N = features.shape
    B_idx, npoint, nsample = idx.shape
    return torch.randn(B, C, npoint, nsample)

def mock_group_points_grad(grad_out, idx, N):
    # grad_out: (B, C, npoint, nsample)
    # output: (B, C, N)
    B, C, npoint, nsample = grad_out.shape
    return torch.randn(B, C, N)

_ext.group_points = mock_group_points
_ext.group_points_grad = mock_group_points_grad

# Create test tensors
B = 2      # batch size
C = 4      # channel dimension
N = 8      # number of points
npoint = 3 # number of centroids
nsample = 5 # number of samples per centroid

features = torch.randn(B, C, N)
idx = torch.randint(0, N, (B, npoint, nsample))

ARGS = {
    "features": features,
    "idx": idx
}
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 GroupingOperation — 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 GroupingOperation 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
MLP_CONV Ran
ran on a synthesized input · constructed an object; no result was computed
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("fc0f3868f0f5e2dc")
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, MLP_CONV: 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_channel": 3,
    "layer_dims": [16, 8, 4],
    "bn": 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 MLP_CONV — 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 MLP_CONV 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
MLP_Res Ran
ran on a synthesized input · constructed an object; no result was computed
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("7e63e6b3d37f1fc5")
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, MLP_Res: 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_dim': 4,
    'hidden_dim': 8,
    'out_dim': 6
}
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 MLP_Res — 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 MLP_Res 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
SeedGenerator Ran
ran on a synthesized input
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("8c6b827a90bafd04")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 3, 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

class MLP_Res(nn.Module):

    def __init__(self, in_dim=128, hidden_dim=None, out_dim=128):
        super(MLP_Res, self).__init__()
        if hidden_dim is None:
            hidden_dim = in_dim
        self.conv_1 = nn.Conv1d(in_dim, hidden_dim, 1)
        self.conv_2 = nn.Conv1d(hidden_dim, out_dim, 1)
        self.conv_shortcut = nn.Conv1d(in_dim, out_dim, 1)

    def forward(self, x):
        """
        Args:
            x: (B, out_dim, n)
        """
        shortcut = self.conv_shortcut(x)
        out = self.conv_2(torch.relu(self.conv_1(x))) + shortcut
        return out

class SeedGenerator(nn.Module):

    def __init__(self, dim_feat=512, num_pc=256):
        super(SeedGenerator, self).__init__()
        self.ps = nn.ConvTranspose1d(dim_feat, 128, num_pc, bias=True)
        self.mlp_1 = MLP_Res(in_dim=dim_feat + 128, hidden_dim=128, out_dim=128)
        self.mlp_2 = MLP_Res(in_dim=128, hidden_dim=64, out_dim=128)
        self.mlp_3 = MLP_Res(in_dim=dim_feat + 128, hidden_dim=128, out_dim=128)
        self.mlp_4 = nn.Sequential(nn.Conv1d(128, 64, 1), nn.ReLU(), nn.Conv1d(64, 3, 1))

    def forward(self, feat):
        """
        Args:
            feat: Tensor (b, dim_feat, 1)
        """
        x1 = self.ps(feat)
        x1 = self.mlp_1(torch.cat([x1, feat.repeat((1, 1, x1.size(2)))], 1))
        x2 = self.mlp_2(x1)
        x3 = self.mlp_3(torch.cat([x2, feat.repeat((1, 1, x2.size(2)))], 1))
        completion = self.mlp_4(x3)
        return completion

INIT_ARGS = {
    'dim_feat': 8,
    'num_pc': 16,
}

ARGS = {
    'feat': torch.randn(2, 8, 1),
}

if __name__ == '__main__':
    model = SeedGenerator(**INIT_ARGS)
    output = model(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Expected shape: (2, 3, 16)")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
Transformer Ran
ran on a synthesized input · constructed an object; no result was computed
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("fd5f8b21108ac85d")
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, Transformer: 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_channel': 8,
    'dim': 16,
    'n_knn': 4,
    'pos_hidden_dim': 8,
    'attn_hidden_multiplier': 2,
}
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 Transformer — 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 Transformer 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
query_knn Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("1825be6622bab350")
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 int, shape (2, 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

torch.manual_seed(0)

# Create sample point clouds
# B=2 (batch size), N=5 (number of source points), C=3 (coordinates)
xyz = torch.randn(2, 5, 3)

# B=2, M=4 (number of query points), C=3
new_xyz = torch.randn(2, 4, 3)

# Number of nearest neighbors to find
nsample = 3

# Whether to include self in the k-NN search
include_self = True

ARGS = {
    "nsample": nsample,
    "xyz": xyz,
    "new_xyz": new_xyz,
    "include_self": include_self,
}
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 int, shape (2, 4, 3) — shape (2, 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
sample_and_group_all Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("225a2727f042aed0")
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)

# Batch size: 2, nsample (number of points): 4, features: 3
batch_size = 2
nsample = 4
num_features = 3

# xyz shape: (B, 3, nsample)
xyz = torch.randn(batch_size, 3, nsample)

# points shape: (B, f, nsample) where f is number of features
points = torch.randn(batch_size, num_features, nsample)

# use_xyz: boolean flag
use_xyz = True

ARGS = {
    "xyz": xyz,
    "points": points,
    "use_xyz": use_xyz,
}
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
Decoder Not yet run
unverified (harvested, no recorded run)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("0a99aea4e840a926")
How it ran

no invocation recorded.

FeatureExtractor Not yet run
unverified (harvested, no recorded run)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("82dceaabbaab2c0a")
How it ran

no invocation recorded.

PointNet_SA_Module_KNN Not yet run
unverified (harvested, no recorded run)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("abb99d2a4bd1907b")
How it ran

no invocation recorded.

SPD Not yet run
unverified (harvested, no recorded run)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("f3ee8ff6841550bf")
How it ran

no invocation recorded.

SkipTransformer Not yet run
unverified (harvested, no recorded run)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("bb507e17130cae13")
How it ran

no invocation recorded.

SnowflakeNet Not yet run
unverified (harvested, no recorded run)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("5d0a9b936733a9c1")
How it ran

no invocation recorded.

fps_subsample Not yet run
unverified (harvested, no recorded run)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("d3d80aea07d1820a")
How it ran

no invocation recorded.

sample_and_group_knn Not yet run
unverified (harvested, no recorded run)
allenxiangx/snowflakenet/models/model_completion.py
code served (permissive licence) · get_code("9eddf08ec6587edd")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 9 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 9 later) and 7 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
sample_and_group_all
brick of this paper's assembly: helper
11 other papers
11 later · 1 only as an assembly brick
2202.09507 (2022-02)
PMP-Net++: Point Cloud Completion by Transformer-Enhanced Multi-step Point Moving Paths
2207.10315 (2022-07)
SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformer
2212.00564 (2022-12)
Leveraging Single-View Images for Unsupervised 3D Point Cloud Completion
and 8 more not named here
GroupingOperation
brick of this paper's assembly: helper
3 other papers
3 later · 3 only as an assembly brick
2207.10315 (2022-07)
SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformer
2303.08240 (2023-03)
Parametric Surface Constrained Upsampler Network for Point Cloud
2401.01552 (2024-01)
CRA-PCN: Point Cloud Completion with Intra-and Inter-level Cross-Resolution Transformers
MLP_CONV
brick of this paper's assembly: helper
3 other papers
3 later · 2 only as an assembly brick
2207.10315 (2022-07)
SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformer
2303.08240 (2023-03)
Parametric Surface Constrained Upsampler Network for Point Cloud
2401.01552 (2024-01)
CRA-PCN: Point Cloud Completion with Intra-and Inter-level Cross-Resolution Transformers
query_knn
brick of this paper's assembly: helper
3 other papers
3 later · 2 only as an assembly brick
2207.10315 (2022-07)
SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformer
2401.01552 (2024-01)
CRA-PCN: Point Cloud Completion with Intra-and Inter-level Cross-Resolution Transformers
2403.05117 (2024-03)
Arbitrary-Scale Point Cloud Upsampling by Voxel-Based Network with Latent Geometric-Consistent Learning
Conv2d
brick of this paper's assembly: helper
2 other papers
2 later · 1 only as an assembly brick
2207.10315 (2022-07)
SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformer
2401.01552 (2024-01)
CRA-PCN: Point Cloud Completion with Intra-and Inter-level Cross-Resolution Transformers
PointNet_SA_Module_KNN
brick of this paper's assembly: helper
2 other papers
2 later · 1 only as an assembly brick
2207.10315 (2022-07)
SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformer
2401.01552 (2024-01)
CRA-PCN: Point Cloud Completion with Intra-and Inter-level Cross-Resolution Transformers
sample_and_group_knn
brick of this paper's assembly: helper
2 other papers
2 later · 2 only as an assembly brick
2207.10315 (2022-07)
SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformer
2401.01552 (2024-01)
CRA-PCN: Point Cloud Completion with Intra-and Inter-level Cross-Resolution Transformers
MLP_Res
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2207.10315 (2022-07)
SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformer
fps_subsample
brick of this paper's assembly: helper
1 other paper
1 later
2312.04060 (2023-12)
Differentiable Registration of Images and LiDAR Point Clouds with VoxelPoint-to-Pixel Matching

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 2026-11-21.

Repositories linked to this paper

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

Abstract

Input (b) TopNet (c) CDN (d) NSFA (e) SnowflakeNet (f) GT * Equal contribution.

For agents

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

get_harvested_code_for_paper("2108.04444")
get_code_for_paper("2108.04444")
have("2108.04444")

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

curl https://syntology.ai/api/ran/2108.04444.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.04444.svg)](https://syntology.ai/paper/2108.04444)

Connect an agent — have() is free.