SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2103.01458 · CVPR · 2021

Diffusion Probabilistic Models for 3D Point Cloud Generation

Wei Hu, Shitong Luo

arXiv · PDF · Open in the Atlas

Code that ran

22 of 29 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (11 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 11 have no contract verdict; 10 failures of our instrument; 7 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/2103.01458.json · badge: /api/ran/2103.01458.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
fpthink/PDGN community
Papers with Code archive
19 ran · 6 no recorded run
luost26/diffusion-point-cloud marked official in Papers with Code
Papers with Code archive
2 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
theamaya/3dlatnav community
Papers with Code archive
0 ran · 1 no recorded run
FunctionStatusWhere it lives
DiffusionPoint Ran
ran on a synthesized input · constructed an object; no result was computed
luost26/diffusion-point-cloud/models/diffusion.py
code served (permissive licence) · get_code("c89046ee64346e9f")
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, DiffusionPoint: 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)

# Define a minimal net that satisfies the interface used in DiffusionPoint
class MinimalNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def forward(self, x, beta, context):
        # x shape: (B, N, d)
        # beta shape: (B,) or scalar
        # context shape: (B, F)
        # Return shape should match x: (B, N, d)
        batch_size, num_points, point_dim = x.size()
        # Simple pass-through network
        out = self.linear(x)
        return out

# Define VarianceSchedule
class VarianceSchedule(nn.Module):
    def __init__(self, num_steps, beta_1, beta_T, mode='linear'):
        super().__init__()
        assert mode in ('linear',)
        self.num_steps = num_steps
        self.beta_1 = beta_1
        self.beta_T = beta_T
        self.mode = mode
        if mode == 'linear':
            betas = torch.linspace(beta_1, beta_T, steps=num_steps)
        betas = torch.cat([torch.zeros([1]), betas], dim=0)
        alphas = 1 - betas
        log_alphas = torch.log(alphas)
        for i in range(1, log_alphas.size(0)):
            log_alphas[i] += log_alphas[i - 1]
        alpha_bars = log_alphas.exp()
        sigmas_flex = torch.sqrt(betas)
        sigmas_inflex = torch.zeros_like(sigmas_flex)
        for i in range(1, sigmas_flex.size(0)):
            sigmas_inflex[i] = (1 - alpha_bars[i - 1]) / (1 - alpha_bars[i]) * betas[i]
        sigmas_inflex = torch.sqrt(sigmas_inflex)
        self.register_buffer('betas', betas)
        self.register_buffer('alphas', alphas)
        self.register_buffer('alpha_bars', alpha_bars)
        self.register_buffer('sigmas_flex', sigmas_flex)
        self.register_buffer('sigmas_inflex', sigmas_inflex)

    def uniform_sample_t(self, batch_size):
        import numpy as np
        ts = np.random.choice(np.arange(1, self.num_steps + 1), batch_size)
        return ts.tolist()

    def get_sigmas(self, t, flexibility):
        assert 0 <= flexibility and flexibility <= 1
        sigmas = self.sigmas_flex[t] * flexibility + self.sigmas_inflex[t] * (1 - flexibility)
        return sigmas

# Create instances
net = MinimalNet()
var_sched = VarianceSchedule(num_steps=5, beta_1=0.0001, beta_T=0.02, mode='linear')

ARGS = {
    'net': net,
    'var_sched': var_sched
}
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 DiffusionPoint — 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 DiffusionPoint 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
PointDiscriminator_1 Ran
ran on a synthesized input · constructed an object; no result was computed
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("25204c1fdea83487")
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, PointDiscriminator_1: 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 PointDiscriminator_1 class expects:
# - num_point: an integer (default 256)
# The forward method expects:
# - x: a tensor of shape (batch_size, 3, num_point)
#   where 3 is the number of channels (x, y, z coordinates)

# Create a small instance for testing
num_point = 16  # Small number of points
batch_size = 2  # Small batch size

# Create input tensor: (batch_size, 3 channels, num_point)
x = torch.randn(batch_size, 3, num_point)

ARGS = {
    "num_point": num_point
}
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 PointDiscriminator_1 — 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 PointDiscriminator_1 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
PointDiscriminator_2 Ran
ran on a synthesized input · constructed an object; no result was computed
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("f9dca246807b4fad")
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, PointDiscriminator_2: 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)

class PointDiscriminator_2(nn.Module):

    def __init__(self, num_point=512):
        super(PointDiscriminator_2, self).__init__()
        self.num_point = num_point
        self.fc1 = nn.Sequential(nn.Conv1d(3, 64, 1), nn.BatchNorm1d(64), nn.LeakyReLU(inplace=True), nn.Conv1d(64, 128, 1), nn.BatchNorm1d(128), nn.LeakyReLU(inplace=True), nn.Conv1d(128, 256, 1), nn.BatchNorm1d(256), nn.LeakyReLU(inplace=True), nn.Conv1d(256, 512, 1), nn.BatchNorm1d(512), nn.LeakyReLU(inplace=True))
        self.maxpool = nn.MaxPool1d(num_point, 1)
        self.mlp = nn.Sequential(nn.Linear(512, 256), nn.LeakyReLU(inplace=True), nn.Linear(256, 64), nn.LeakyReLU(inplace=True), nn.Linear(64, 1))

    def forward(self, x):
        batchsize = x.size()[0]
        x1 = self.fc1(x)
        x2 = self.maxpool(x1)
        x2 = x2.view(batchsize, 512)
        x3 = self.mlp(x2)
        return x3

ARGS = {
    "num_point": 4
}
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 PointDiscriminator_2 — 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 PointDiscriminator_2 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
PointDiscriminator_3 Ran
ran on a synthesized input · constructed an object; no result was computed
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("35cbd79bd975fc0c")
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, PointDiscriminator_3: 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)

class PointDiscriminator_3(nn.Module):

    def __init__(self, num_point=1024):
        super(PointDiscriminator_3, self).__init__()
        self.num_point = num_point
        self.fc1 = nn.Sequential(nn.Conv1d(3, 64, 1), nn.BatchNorm1d(64), nn.LeakyReLU(inplace=True), nn.Conv1d(64, 128, 1), nn.BatchNorm1d(128), nn.LeakyReLU(inplace=True), nn.Conv1d(128, 256, 1), nn.BatchNorm1d(256), nn.LeakyReLU(inplace=True), nn.Conv1d(256, 512, 1), nn.BatchNorm1d(512), nn.LeakyReLU(inplace=True))
        self.maxpool = nn.MaxPool1d(num_point, 1)
        self.mlp = nn.Sequential(nn.Linear(512, 256), nn.LeakyReLU(inplace=True), nn.Linear(256, 64), nn.LeakyReLU(inplace=True), nn.Linear(64, 1))

    def forward(self, x):
        batchsize = x.size()[0]
        x1 = self.fc1(x)
        x2 = self.maxpool(x1)
        x2 = x2.view(batchsize, 512)
        x3 = self.mlp(x2)
        return x3

ARGS = {
    'num_point': 4
}

model = PointDiscriminator_3(**ARGS)
x = torch.randn(2, 3, 4)
output = model(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 PointDiscriminator_3 — 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 PointDiscriminator_3 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
PointDiscriminator_4 Ran
ran on a synthesized input · constructed an object; no result was computed
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("88b6db90221f3084")
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, PointDiscriminator_4: 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 PointDiscriminator_4 class expects:
# - num_point: int parameter for the number of points (default 2048)
# - forward() expects input x with shape (batch_size, 3, num_point)

# Create a small point cloud input tensor
# Shape: (batch_size=2, channels=3, num_points=4)
batch_size = 2
num_channels = 3
num_points = 4

x = torch.randn(batch_size, num_channels, num_points)

ARGS = {
    "num_point": num_points,
}
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 PointDiscriminator_4 — 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 PointDiscriminator_4 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
ShapeNetCore Ran
ran on a synthesized input · constructed an object; no result was computed
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("6e626337710c79fd")
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, ShapeNetCore: 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
import tempfile
import os
import h5py

torch.manual_seed(0)

# Create a temporary HDF5 file with minimal ShapeNet-like structure
temp_dir = tempfile.mkdtemp()
h5_path = os.path.join(temp_dir, 'test_dataset.h5')

# Create minimal HDF5 file with required structure
with h5py.File(h5_path, 'w') as f:
    # Add data for a couple of categories
    for synsetid in ['02691156', '03001627']:  # airplane, chair
        cat_group = f.create_group(synsetid)
        for split in ['train', 'val', 'test']:
            # Create dummy point clouds: shape (5, 1024, 3) for 5 samples
            data = np.random.randn(5, 1024, 3).astype(np.float32)
            cat_group.create_dataset(split, data=data)

ARGS = {
    'cates_list': 'airplane',  # single category as string (will be wrapped in list by __init__)
    'split': 'train',
    'scale_mode': 'global_unit',
    'path': h5_path,
    'transform': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
h5py, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ShapeNetCore — the run constructed an instance of its own class and did not call it; no shape is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
VarianceSchedule Ran
ran on a synthesized input · constructed an object; no result was computed
luost26/diffusion-point-cloud/models/diffusion.py
code served (permissive licence) · get_code("380bb5e9ac8495db")
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, VarianceSchedule: 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

torch.manual_seed(0)
np.random.seed(0)

ARGS = {
    'num_steps': 10,
    'beta_1': 0.0001,
    'beta_T': 0.02,
    'mode': 'linear'
}
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 VarianceSchedule — 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 VarianceSchedule 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
bilateral_block_l1 Ran
ran on a synthesized input · constructed an object; no result was computed
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("6c87f76cd845e511")
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, bilateral_block_l1: 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)

# Test input tensor
# Shape: [B, Fin, N] where B=2 (batch), Fin=4 (input features), N=8 (points)
x = torch.randn(2, 4, 8)

ARGS = {
    "Fin": 4,
    "Fout": 6,
    "maxpool": 4,
    "stride": 1,
    "num_k": 20
}
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 bilateral_block_l1 — 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 bilateral_block_l1 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
bilateral_upsample_edgeConv Ran
ran on a synthesized input · constructed an object; no result was computed
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("cde1974d0e8c3464")
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, bilateral_upsample_edgeConv: 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)

# Parameters for bilateral_upsample_edgeConv
Fin = 4
Fout = 8
k = 6
num = -1
softmax = True

ARGS = {
    'Fin': Fin,
    'Fout': Fout,
    'k': k,
    'num': num,
    'softmax': softmax
}
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 bilateral_upsample_edgeConv — 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 bilateral_upsample_edgeConv 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
conv2dbr Ran
ran on a synthesized input · constructed an object; no result was computed
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("8617c57ff67112bf")
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, conv2dbr: 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 = {
    "Fin": 3,
    "Fout": 16,
    "kernel_size": 3,
    "stride": 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 conv2dbr — 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 conv2dbr 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_edge_features Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("1e042976d61e8c8a")
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, 8, 5, 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)

# Construct arguments for get_edge_features
# x: point cloud [B, dims, N]
# k: kNN neighbours
# num: optional parameter (default -1)

B = 2  # batch size
dims = 4  # dimensions
N = 5  # number of points

# Create a small point cloud tensor [B, dims, N]
x = torch.randn(B, dims, N)

# Set k to a reasonable value for kNN (must be < N to get k+1 neighbors)
k = 3

# num parameter defaults to -1
num = -1

ARGS = {
    "x": x,
    "k": k,
    "num": num,
}
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, 8, 5, 3) — shape (2, 8, 5, 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
get_edge_features_xyz Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("096e5ea95c5f7eb6")
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)

# Create minimal test tensors
B = 2  # batch size
dims = 4  # feature dimensions
N = 5  # number of points
k = 3  # number of neighbors

# x: point cloud features [B, dims, N]
x = torch.randn(B, dims, N)

# pc: point cloud coordinates [B, 3, N]
pc = torch.randn(B, 3, N)

# k: number of neighbors (scalar)
k_val = k

# num: optional parameter (default -1)
num = -1

ARGS = {
    "x": x,
    "pc": pc,
    "k": k_val,
    "num": num
}
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
jitter_point_cloud Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("bdd7784c7336e843")
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 (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 numpy as np

np.random.seed(0)

# Create a small batch of point clouds
# B=2 (batch size), N=4 (number of points), C=3 (coordinates: x, y, z)
batch_data = np.random.randn(2, 4, 3).astype(np.float32)

ARGS = {
    "batch_data": batch_data,
    "sigma": 0.01,
    "clip": 0.05,
}
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 (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
pc_normalize Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from aboulch/ConvPoint
pointer only; the stored copy get_code serves, from aboulch/ConvPoint, is pointer only (licence: NOASSERTION) · get_code("ec413739d406e611")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

random_scale_point_cloud Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("c4bfd74b0f6a7d8d")
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 (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 numpy as np

np.random.seed(0)

# Create a small batch of point clouds
# B=2 (batch size), N=4 (number of points), C=3 (coordinates x,y,z)
batch_data = np.random.randn(2, 4, 3).astype(np.float32)

ARGS = {
    'batch_data': batch_data,
    'scale_low': 0.8,
    'scale_high': 1.25,
}
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 (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
rotate_perturbation_point_cloud Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("79446b78c27526f0")
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 (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 numpy as np

# Create a small batch of point clouds
# BxNx3 array where B=2 (batch size), N=4 (number of points), 3 (x,y,z coordinates)
batch_data = np.random.randn(2, 4, 3).astype(np.float32)

ARGS = {
    "batch_data": batch_data,
    "angle_sigma": 0.06,
    "angle_clip": 0.18,
}
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 (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
rotate_perturbation_point_cloud_with_normal Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("3db497496c6e4cc0")
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 (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 numpy as np

# Create batch_data: BxNx6 array where B=batch_size, N=num_points, 6=(x,y,z,nx,ny,nz)
# B=2 (batch size), N=4 (num points), 6 (coordinates + normals)
batch_size = 2
num_points = 4
batch_data = np.random.randn(batch_size, num_points, 6).astype(np.float32)

ARGS = {
    "batch_data": batch_data,
    "angle_sigma": 0.06,
    "angle_clip": 0.18,
}
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 (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
rotate_point_cloud Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("adb85367d5bf360a")
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 (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 numpy as np

np.random.seed(0)

# Create a small batch of point clouds
# BxNx3 array where B=2 (batch size), N=4 (number of points), 3 (x,y,z coordinates)
batch_data = np.random.randn(2, 4, 3).astype(np.float32)

ARGS = {
    "batch_data": batch_data
}
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 (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
rotate_point_cloud_with_normal Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("9b39b2ae031d10f6")
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 (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 numpy as np

# Create a small batch of point clouds with XYZ and normal coordinates
# Shape: (B, N, 6) where B=batch size, N=num points, 6=3 for XYZ + 3 for normal
batch_size = 2
num_points = 4
feature_dim = 6

batch_xyz_normal = np.random.randn(batch_size, num_points, feature_dim).astype(np.float32)

ARGS = {
    "batch_xyz_normal": batch_xyz_normal
}
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 (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
shift_point_cloud Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("265977b03f1bbb45")
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 (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 numpy as np

np.random.seed(0)

# Create a small batch of point clouds
# B=2 (batch size), N=4 (number of points), C=3 (coordinates x,y,z)
batch_data = np.random.randn(2, 4, 3).astype(np.float32)

ARGS = {
    "batch_data": batch_data,
    "shift_range": 0.1,
}
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 (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
shuffle_points Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("a97f76b9c0811c93")
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 (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 numpy as np

# Create a small batch of point cloud data
# BxNxC format: B=batch size, N=number of points, C=channels (coordinates)
batch_size = 2
num_points = 4
channels = 3

batch_data = np.random.randn(batch_size, num_points, channels).astype(np.float32)

ARGS = {
    "batch_data": batch_data
}
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 (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
upsample_edgeConv Ran
ran on a synthesized input · constructed an object; no result was computed
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("7f1525d96299a288")
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, upsample_edgeConv: 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 ARGS dict with all parameters for upsample_edgeConv
ARGS = {
    'Fin': 4,
    'Fout': 8,
    'k': 6,
    'num': -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 upsample_edgeConv — 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 upsample_edgeConv 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
DiffusionPoint Not yet run
unverified (harvested, no recorded run)
theamaya/3dlatnav/models/Diffusion/models/diffusion.py
pointer only (licence: NONE) · get_code("1e48b52609d7ac4f")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, theamaya/3dlatnav when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

ModelNetDataset Not yet run
unverified (harvested, no recorded run)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("96a47b47754cc0a4")
How it ran

no invocation recorded.

PDGNet Not yet run
unverified (harvested, no recorded run)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("0c0259c3ed40c727")
How it ran

no invocation recorded.

PointGenerator Not yet run
unverified (harvested, no recorded run)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("4e304106a8e2b1df")
How it ran

no invocation recorded.

bilateral_block_l2 Not yet run
unverified (harvested, no recorded run)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("f29c498243b560a7")
How it ran

no invocation recorded.

bilateral_block_l3 Not yet run
unverified (harvested, no recorded run)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("f5e60736bb09fc86")
How it ran

no invocation recorded.

bilateral_block_l4 Not yet run
unverified (harvested, no recorded run)
fpthink/PDGN/models/PDGNet.py
code served (permissive licence) · get_code("1bfe4165bf8b09b1")
How it ran

no invocation recorded.

Also present in other papers' code

Of 29 samples lifted from this paper's linked repositories, 29 carry a function identity and 0 do not. Of the 29 distinct functions with an identity, 5 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 2 the same month, 5 later) and 24 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
pc_normalize
brick of this paper's assembly: helper
36 other papers
7 dated earlier, 1 the same month, 28 later
1803.05827 (2018-03)
Local Spectral Graph Convolution for Point Set Feature Learning
1904.02375 (2019-04)
ConvPoint: Continuous Convolutions for Point Cloud Processing
1904.10014 (2019-04)
Linked Dynamic Graph CNN: Learning on Point Cloud via Linking Hierarchical Features
and 33 more not named here
shuffle_points
brick of this paper's assembly: helper
11 other papers
3 dated earlier, 1 the same month, 7 later
1811.07246 (2018-11)
PointConv: Deep Convolutional Networks on 3D Point Clouds
1812.11647 (2018-12)
Path-Invariant Map Networks
2006.14865 (2020-06)
RPM-Net: Recurrent Prediction of Motion and Parts from Point Cloud
and 8 more not named here
rotate_point_cloud
brick of this paper's assembly: helper
7 other papers
4 dated earlier, 3 later
1803.11527 (2018-03)
SpiderCNN: Deep Learning on Point Sets with Parameterized Convolutional Filters
1904.10014 (2019-04)
Linked Dynamic Graph CNN: Learning on Point Cloud via Linking Hierarchical Features
1908.04616 (2019-08)
Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Data
and 4 more not named here
rotate_perturbation_point_cloud
brick of this paper's assembly: helper
2 other papers
2 later
2206.00927 (2022-06)
DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps
Nunes_Temporal_Consistent_3D_LiDAR_Representation_Learning_for_Semantic_Perception_in_CVPR_2023_paper (2023 (year only))
Temporal Consistent 3D LiDAR Representation Learning for Semantic Perception in Autonomous Driving
jitter_point_cloud
brick of this paper's assembly: helper
1 other paper
1 later
Nunes_Temporal_Consistent_3D_LiDAR_Representation_Learning_for_Semantic_Perception_in_CVPR_2023_paper (2023 (year only))
Temporal Consistent 3D LiDAR Representation Learning for Semantic Perception in Autonomous Driving

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

We present a probabilistic model for point cloud generation, which is fundamental for various 3D vision tasks such as shape completion, upsampling, synthesis and data augmentation. Inspired by the diffusion process in nonequilibrium thermodynamics, we view points in point clouds as particles in a thermodynamic system in contact with a heat bath, which diffuse from the original distribution to a noise distribution. Point cloud generation thus amounts to learning the reverse diffusion process that transforms the noise distribution to the distribution of a desired shape. Specifically, we propose to model the reverse diffusion process for point clouds as a Markov chain conditioned on certain shape latent. We derive the variational bound in closed form for training and provide implementations of the model. Experimental results demonstrate that our model achieves competitive performance in point cloud generation and auto-encoding. The code is available at https://github.com/luost26/diffusionpoint-cloud.

For agents

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

get_harvested_code_for_paper("2103.01458")
get_code_for_paper("2103.01458")
have("2103.01458")

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

curl https://syntology.ai/api/ran/2103.01458.json

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

[![Syntology run record](https://syntology.ai/api/ran/2103.01458.svg)](https://syntology.ai/paper/2103.01458)

Connect an agent — have() is free.