SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.03746 · 2021

Unsupervised Learning of Fine Structure Generation for 3D Point Clouds by 2D Projection Matching

arXiv · PDF · Open in the Atlas

Code that ran

9 of 11 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 9 have no contract verdict; 0 failures 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/2108.03746.json · badge: /api/ran/2108.03746.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
chenchao15/2d_projection_matching canonical (classifier label)
classifier basis: census
9 ran · 2 no recorded run
FunctionStatusWhere it lives
drc_loss Ran
ran on a synthesized input
chenchao15/2d_projection_matching/2Dpm/util/losses.py
code served (permissive licence) · get_code("60ee93b1c66fab31")
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 EagerTensor, 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 tensorflow as tf
import numpy as np

class Config:
    def __init__(self):
        self.vox_size = 2

cfg = Config()

# Create small tensors for the function
# probs shape: [vox_size+1, batch, height, width, depth] = [3, 2, 4, 5, 3]
probs = tf.constant(
    np.random.rand(3, 2, 4, 5, 3).astype(np.float32),
    dtype=tf.float32
)

# gt_proj shape: [batch, height, width, depth] = [2, 4, 5, 3]
gt_proj = tf.constant(
    np.random.rand(2, 4, 5, 3).astype(np.float32),
    dtype=tf.float32
)

ARGS = {
    "cfg": cfg,
    "probs": probs,
    "gt_proj": gt_proj
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
edgeLength Ran
ran on a synthesized input
chenchao15/2d_projection_matching/2Dpm/densify/utils.py
code served (permissive licence) · get_code("a69a6eb516926e21")
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 float64, ndim 0, 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 numpy as np

# V: array of vertex coordinates
# E: array of edge definitions (pairs of vertex indices)
# i: index of the edge to measure

# Create a small set of vertices in 3D space
V = np.array([
    [0.0, 0.0, 0.0],
    [1.0, 0.0, 0.0],
    [0.0, 1.0, 0.0],
    [1.0, 1.0, 1.0]
])

# Create edges as pairs of vertex indices
E = np.array([
    [0, 1],
    [1, 2],
    [2, 3],
    [0, 3]
])

# Index of edge to measure
i = 0

ARGS = {
    "V": V,
    "E": E,
    "i": i
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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_dropout_prob Ran
ran on a synthesized input
chenchao15/2d_projection_matching/2Dpm/models/models.py
code served (permissive licence) · get_code("8df602630b1c437a")
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 EagerTensor, 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 tensorflow as tf

class Config:
    def __init__(self):
        self.pc_point_dropout_scheduled = True
        self.pc_point_dropout = 0.1
        self.pc_point_dropout_exponential_schedule = False
        self.max_number_of_steps = 1000
        self.pc_point_dropout_start_step = 100
        self.pc_point_dropout_end_step = 900

cfg = Config()
global_step = tf.constant(500, dtype=tf.int32)

ARGS = {
    "cfg": cfg,
    "global_step": global_step
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
get_smooth_sigma Ran
ran on a synthesized input
chenchao15/2d_projection_matching/2Dpm/models/models.py
code served (permissive licence) · get_code("6e6dd4087fa03641")
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 EagerTensor, ndim 0, 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 tensorflow as tf

class Config:
    def __init__(self):
        self.max_number_of_steps = 100
        self.pc_relative_sigma_end = 0.5
        self.pc_relative_sigma = 0.1

cfg = Config()
global_step = 25

ARGS = {
    "cfg": cfg,
    "global_step": global_step
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 0, dtype family float — no output shape is stated: observe_signature.py could not observe it (import:TypeError)
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
normalise_depthmap Ran
ran on a synthesized input
chenchao15/2d_projection_matching/2Dpm/main/predict.py
code served (permissive licence) · get_code("dade09e8e44b9b60")
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 (2, 4) 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 sample depth map as a numpy array
# The function expects a numpy array with depth values
# Using a small array with values that will be clipped to [1.5, 2.5] range
depth_map = np.array([
    [1.0, 2.0, 3.0, 2.5],
    [1.5, 2.2, 0.5, 2.8]
], dtype=np.float32)

ARGS = {
    "depth_map": depth_map
}
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 (2, 4) — shape (2, 4) 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
parseObj Ran
ran on a synthesized input
chenchao15/2d_projection_matching/2Dpm/densify/utils.py
code served (permissive licence) · get_code("5db33295fea6d2a3")
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
import tempfile
import os

# Create a temporary OBJ file with valid geometry
temp_obj_content = """# Simple triangle mesh
v 0.0 0.0 0.0
v 1.0 0.0 0.0
v 0.0 1.0 0.0
v 1.0 1.0 0.0
v 0.0 0.0 1.0
f 1/1/1 2/2/1 3/3/1
f 2/2/1 4/4/1 3/3/1
f 1/1/1 2/2/1 5/5/1
"""

# Write to temporary file
with tempfile.NamedTemporaryFile(mode='w', suffix='.obj', delete=False) as f:
    f.write(temp_obj_content)
    temp_fname = f.name

try:
    # Define ARGS mapping the parameter to the temporary file path
    ARGS = {
        'fname': temp_fname
    }
finally:
    # Clean up will happen after the function is called
    pass
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
pool_single_view Ran
ran on a synthesized input
chenchao15/2d_projection_matching/2Dpm/data_preprocessor/data_preprocessor.py
code served (permissive licence) · get_code("dd16754ef3053ae3")
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 EagerTensor, ndim 2, dtype family float, shape (2, 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 tensorflow as tf

class Config:
    def __init__(self, batch_size, step_size):
        self.batch_size = batch_size
        self.step_size = step_size

cfg = Config(batch_size=2, step_size=4)

tensor = tf.constant([
    [1.0, 2.0, 3.0],
    [4.0, 5.0, 6.0],
    [7.0, 8.0, 9.0],
    [10.0, 11.0, 12.0],
    [13.0, 14.0, 15.0],
    [16.0, 17.0, 18.0],
    [19.0, 20.0, 21.0],
    [22.0, 23.0, 24.0],
])

view_idx = 1

ARGS = {
    "cfg": cfg,
    "tensor": tensor,
    "view_idx": view_idx,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (2, 3) — shape (2, 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
removeWeirdDuplicate Ran
ran on a synthesized input
chenchao15/2d_projection_matching/2Dpm/densify/utils.py
code served (permissive licence) · get_code("d9800bc5370d0352")
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 2, dtype family int; 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 numpy as np

torch.manual_seed(0)

# The function expects F to be a 2D array-like object that:
# 1. Supports .sort(axis=1) - so it should be a numpy array
# 2. Can be iterated to create a list
# 3. Elements are indexable with [0], [1], [2]
# This suggests F should be a numpy array with shape (N, 3) or similar

# Create a small numpy array with some duplicate rows to test the deduplication
F = np.array([
    [1, 2, 3],
    [1, 2, 3],
    [2, 1, 4],
    [1, 2, 3],
    [2, 1, 4],
    [3, 0, 1]
], dtype=np.int32)

ARGS = {
    'F': F
}
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 list, ndim 2, dtype family int — 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
tf_repeat_0 Ran
ran on a synthesized input
chenchao15/2d_projection_matching/2Dpm/models/models.py
code served (permissive licence) · get_code("5614bddbbf44e931")
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 EagerTensor, ndim 3, dtype family float, shape (10, 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 tensorflow as tf
import numpy as np

# Create a small input tensor for testing
# Using shape (2, 4, 3) - batch size 2, 4 features, 3 dimensions
input_tensor = tf.constant(np.random.randn(2, 4, 3), dtype=tf.float32)

# Set num to repeat along the expanded dimension
num_repeats = 5

ARGS = {
    "input": input_tensor,
    "num": num_repeats
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 3, dtype family float, shape (10, 4, 3) — shape (10, 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
drc_rgb_loss Not yet run
unverified (harvested, no recorded run)
chenchao15/2d_projection_matching/2Dpm/util/losses.py
code served (permissive licence) · get_code("c8f563881aa98dd4")
How it ran

no invocation recorded.

regularization_loss Not yet run
unverified (harvested, no recorded run)
chenchao15/2d_projection_matching/2Dpm/util/losses.py
code served (permissive licence) · get_code("eecdb72ca340d002")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 7 are also present in other papers whose code we hold (7 with at least one paper dated earlier, 0 the same month, 0 later) and 4 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
drc_loss 1 other paper
1 dated earlier
1810.09381 (2018-10)
Unsupervised Learning of Shape and Pose with Differentiable Point Clouds
drc_rgb_loss 1 other paper
1 dated earlier
1810.09381 (2018-10)
Unsupervised Learning of Shape and Pose with Differentiable Point Clouds
get_dropout_prob 1 other paper
1 dated earlier
1810.09381 (2018-10)
Unsupervised Learning of Shape and Pose with Differentiable Point Clouds
get_smooth_sigma 1 other paper
1 dated earlier
1810.09381 (2018-10)
Unsupervised Learning of Shape and Pose with Differentiable Point Clouds
pool_single_view 1 other paper
1 dated earlier
1810.09381 (2018-10)
Unsupervised Learning of Shape and Pose with Differentiable Point Clouds
regularization_loss 1 other paper
1 dated earlier
1810.09381 (2018-10)
Unsupervised Learning of Shape and Pose with Differentiable Point Clouds
tf_repeat_0 1 other paper
1 dated earlier
1810.09381 (2018-10)
Unsupervised Learning of Shape and Pose with Differentiable Point Clouds

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

Learning to generate 3D point clouds without 3D supervision is an important but challenging problem. Current solutions leverage various differentiable renderers to project the generated 3D point clouds onto a 2D image plane, and train deep neural networks using the per-pixel difference with 2D ground truth images. However, these solutions are still struggling to fully recover fine structures of 3D shapes, such as thin tubes or planes. To resolve this issue, we propose an unsupervised approach for 3D point cloud generation with fine structures. Specifically, we cast 3D point cloud learning as a 2D projection matching problem. Rather than using entire 2D silhouette images as a regular pixel supervision, we introduce structure adaptive sampling to randomly sample 2D points within the silhouettes as an irregular point supervision, which alleviates the consistency issue of sampling from different view angles. Our method pushes the neural network to generate a 3D point cloud whose 2D projections match the irregular point supervision from different view angles. Our 2D projection matching approach enables the neural network to learn more accurate structure information than using the per-pixel difference, especially for fine and thin 3D structures. Our method can recover fine 3D structures from 2D silhouette images at different resolutions, and is robust to different sampling methods and point number in irregular point supervision. Our method outperforms others under widely used benchmarks. Our code, data and models are available at https://github.com/chenchao15/2D\_projection\_matching.

For agents

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

get_harvested_code_for_paper("2108.03746")
get_code_for_paper("2108.03746")
have("2108.03746")

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

curl https://syntology.ai/api/ran/2108.03746.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.03746.svg)](https://syntology.ai/paper/2108.03746)

Connect an agent — have() is free.