SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.10314 · CVPR · 2022

Voxel Set Transformer: A Set-to-Set Approach to 3D Object Detection from Point Clouds

Lei Zhang, Shuai Li, Chenhang He, Ruihuang Li, Drop Pad

arXiv · PDF · Open in the Atlas

Code that ran

4 of 6 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 4 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/2203.10314.json · badge: /api/ran/2203.10314.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
skyhehe123/VoxSeT canonical (classifier label)
classifier basis not recorded on the link
4 ran · 2 no recorded run
FunctionStatusWhere it lives
cfg_from_yaml_file Ran
ran; our contract draft was wrong, not the code
skyhehe123/VoxSeT/pcdet/config.py
code served (permissive licence) · get_code("44db2351bcc0bffe")
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 EasyDict. 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 yaml
from easydict import EasyDict
import tempfile
import os

# Create temporary YAML files for testing
temp_dir = tempfile.mkdtemp()

# Create a base config file
base_config_path = os.path.join(temp_dir, 'base_config.yaml')
with open(base_config_path, 'w') as f:
    yaml.dump({
        'model': {
            'layers': 3,
            'hidden_dim': 64
        },
        'learning_rate': 0.001,
        'batch_size': 32
    }, f)

# Create main config file that references base config
main_config_path = os.path.join(temp_dir, 'main_config.yaml')
with open(main_config_path, 'w') as f:
    yaml.dump({
        '_BASE_CONFIG_': base_config_path,
        'model': {
            'layers': 5
        },
        'epochs': 100,
        'learning_rate': 0.01
    }, f)

# Initialize config as EasyDict
initial_config = EasyDict()

ARGS = {
    'cfg_file': main_config_path,
    'config': initial_config
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
easydict, yaml — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EasyDict — a EasyDict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a EasyDict 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
compute_fg_mask Ran
ran; our contract draft was wrong, not the code
skyhehe123/VoxSeT/pcdet/utils/loss_utils.py
code served (permissive licence) · get_code("65fe32ede00e7dcb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (2, 128, 256) 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)

# gt_boxes2d: (B, N, 4) tensor with 2D box coordinates [u1, v1, u2, v2]
# B=2 (batch size), N=3 (number of boxes), 4 coordinates per box
gt_boxes2d = torch.tensor([
    [[10.5, 20.5, 50.5, 60.5],
     [5.0, 10.0, 30.0, 40.0],
     [70.0, 80.0, 100.0, 120.0]],
    [[15.5, 25.5, 55.5, 65.5],
     [8.0, 12.0, 32.0, 42.0],
     [75.0, 85.0, 105.0, 125.0]]
], dtype=torch.float32)

# shape: desired output mask shape (B, H, W) = (2, 128, 256)
shape = torch.Size([2, 128, 256])

# downsample_factor: integer downsampling factor
downsample_factor = 2

# device: torch device (CPU only)
device = torch.device('cpu')

ARGS = {
    'gt_boxes2d': gt_boxes2d,
    'shape': shape,
    'downsample_factor': downsample_factor,
    'device': device
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 128, 256) — shape (2, 128, 256) 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
merge_new_config Ran
ran; our contract draft was wrong, not the code
skyhehe123/VoxSeT/pcdet/config.py
code served (permissive licence) · get_code("50e8e8cdfc5129f0")
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 EasyDict. 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 yaml
from easydict import EasyDict
import tempfile
import os

# Create a temporary base config file
temp_dir = tempfile.mkdtemp()
base_config_path = os.path.join(temp_dir, 'base_config.yaml')

base_config_content = """
model:
  type: simple
  layers: 3
learning_rate: 0.001
batch_size: 32
"""

with open(base_config_path, 'w') as f:
    f.write(base_config_content)

# Create the base config as an EasyDict
config = EasyDict({
    'model': EasyDict({'type': 'simple', 'layers': 3}),
    'learning_rate': 0.001,
    'batch_size': 32
})

# Create the new_config with _BASE_CONFIG_ pointing to our temp file
new_config = {
    '_BASE_CONFIG_': base_config_path,
    'learning_rate': 0.01,
    'model': {
        'layers': 5
    },
    'optimizer': 'adam'
}

ARGS = {
    'config': config,
    'new_config': new_config
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
easydict, yaml — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EasyDict — a EasyDict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a EasyDict 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
neg_loss_cornernet Ran
ran; our contract draft was wrong, not the code
skyhehe123/VoxSeT/pcdet/utils/loss_utils.py
code served (permissive licence) · get_code("488b91d67a807558")
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 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# pred: (batch x c x h x w) - predictions, should be probabilities in (0, 1)
# gt: (batch x c x h x w) - ground truth, values in [0, 1]
# mask: (batch x h x w) - optional spatial mask

batch_size = 2
num_channels = 4
height = 3
width = 5

# Create pred tensor - predictions (batch x c x h x w)
# Use values between 0.1 and 0.9 to avoid log(0) or log(1) issues
pred = torch.rand(batch_size, num_channels, height, width) * 0.8 + 0.1

# Create gt tensor - ground truth with mostly 0s and some 1s (batch x c x h x w)
gt = torch.zeros(batch_size, num_channels, height, width)
gt[0, 0, 0, 0] = 1.0
gt[1, 2, 1, 2] = 1.0

# Create mask tensor - spatial mask (batch x h x w)
mask = torch.ones(batch_size, height, width)
mask[0, 0, 0] = 0.0  # mask out one position

ARGS = {
    "pred": pred,
    "gt": gt,
    "mask": mask,
}
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 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_corner_loss_lidar Not yet run
unverified (harvested, no recorded run)
skyhehe123/VoxSeT/pcdet/utils/loss_utils.py
code served (permissive licence) · get_code("1780d388cc532a6d")
How it ran

no invocation recorded.

post_act_block Not yet run
unverified (harvested, no recorded run)
skyhehe123/VoxSeT/pcdet/models/backbones_3d/spconv_backbone.py
code served (permissive licence) · get_code("2b4e0558df870bf9")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 6 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 2 the same month, 6 later) and 0 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
get_corner_loss_lidar 49 other papers
8 dated earlier, 2 the same month, 39 later
1907.03670 (2019-07)
From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network
1912.13192 (2019-12)
PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection
2005.10821 (2020-05)
Hierarchical Multi-Scale Attention for Semantic Segmentation
and 46 more not named here
cfg_from_yaml_file 37 other papers
2 dated earlier, 35 later
1907.03670 (2019-07)
From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network
2104.11896 (2021-04)
M3DeTR: Multi-representation, Multi-scale, Mutual-relation 3D Object Detection with Transformers
2204.06577 (2022-04)
OccAM's Laser: Occlusion-based Attribution Maps for 3D Object Detectors on LiDAR Data
and 34 more not named here
compute_fg_mask 35 other papers
3 dated earlier, 1 the same month, 31 later
1907.03670 (2019-07)
From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network
2104.11896 (2021-04)
M3DeTR: Multi-representation, Multi-scale, Mutual-relation 3D Object Detection with Transformers
2105.00373 (2021-05)
Investigating the Impact of Multi-LiDAR Placement on Object Detection for Autonomous Driving
and 32 more not named here
merge_new_config 35 other papers
2 dated earlier, 33 later
1907.03670 (2019-07)
From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network
2104.11896 (2021-04)
M3DeTR: Multi-representation, Multi-scale, Mutual-relation 3D Object Detection with Transformers
2204.06577 (2022-04)
OccAM's Laser: Occlusion-based Attribution Maps for 3D Object Detectors on LiDAR Data
and 32 more not named here
neg_loss_cornernet 32 other papers
2 dated earlier, 30 later
1907.03670 (2019-07)
From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network
2104.11896 (2021-04)
M3DeTR: Multi-representation, Multi-scale, Mutual-relation 3D Object Detection with Transformers
2204.06577 (2022-04)
OccAM's Laser: Occlusion-based Attribution Maps for 3D Object Detectors on LiDAR Data
and 29 more not named here
post_act_block 28 other papers
2 dated earlier, 26 later
1907.03670 (2019-07)
From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network
2104.11896 (2021-04)
M3DeTR: Multi-representation, Multi-scale, Mutual-relation 3D Object Detection with Transformers
2204.06577 (2022-04)
OccAM's Laser: Occlusion-based Attribution Maps for 3D Object Detectors on LiDAR Data
and 25 more not named here

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

Transformer has demonstrated promising performance in many 2D vision tasks. However, it is cumbersome to compute the self-attention on large-scale point cloud data because point cloud is a long sequence and unevenly distributed in 3D space. To solve this issue, existing methods usually compute self-attention locally by grouping the points into clusters of the same size, or perform convolutional self-attention on a discretized representation. However, the former results in stochastic point dropout, while the latter typically has narrow attention fields. In this paper, we propose a novel voxel-based architecture, namely Voxel Set Transformer (VoxSeT), to detect 3D objects from point clouds by means of set-to-set translation. VoxSeT is built upon a voxel-based set attention (VSA) module, which reduces the self-attention in each voxel by two crossattentions and models features in a hidden space induced by a group of latent codes. With the VSA module, VoxSeT can manage voxelized point clusters with arbitrary size in a wide range, and process them in parallel with linear complexity. The proposed VoxSeT integrates the high performance of transformer with the efficiency of voxelbased model, which can be used as a good alternative to the convolutional and point-based backbones. VoxSeT reports competitive results on the KITTI and Waymo detection benchmarks. The source codes can be found at https://github.com/skyhehe123/VoxSeT.

For agents

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

get_harvested_code_for_paper("2203.10314")
get_code_for_paper("2203.10314")
have("2203.10314")

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

curl https://syntology.ai/api/ran/2203.10314.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.10314.svg)](https://syntology.ai/paper/2203.10314)

Connect an agent — have() is free.