SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.07300 · 2021

Learning to Estimate Robust 3D Human Mesh from In-the-Wild Crowded Scenes

arXiv · PDF · Open in the Atlas

Code that ran

8 of 9 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 5 have no contract verdict; 3 failures of our instrument; 1 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/2104.07300.json · badge: /api/ran/2104.07300.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
hongsukchoi/3dcrowdnet_release canonical (classifier label)
classifier basis not recorded on the link
8 ran · 1 no recorded run
FunctionStatusWhere it lives
bbox_iou Ran
ran on a synthesized input
hongsukchoi/3dcrowdnet_release/tool/match_3dpw_2dpose.py
code served (permissive licence) · get_code("fbb3863e95291441")
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 1, dtype family float, shape (2,) 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 two bounding boxes tensors
# Format: [x1, y1, x2, y2] for each box
# box1: 2 bounding boxes
box1 = torch.tensor([
    [10.0, 20.0, 50.0, 60.0],
    [15.0, 25.0, 55.0, 65.0]
], dtype=torch.float32)

# box2: 2 bounding boxes
box2 = torch.tensor([
    [30.0, 40.0, 70.0, 80.0],
    [20.0, 30.0, 60.0, 70.0]
], dtype=torch.float32)

ARGS = {
    "box1": box1,
    "box2": box2
}
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 1, dtype family float, shape (2,) — shape (2,) 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
compute_CrowdIndex Ran
ran on a synthesized input
hongsukchoi/3dcrowdnet_release/tool/check_crowdidx.py
code served (permissive licence) · get_code("6671d7cae5239d75")
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 float, 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
import numpy as np

torch.manual_seed(0)

# ref_bbox: a bounding box in format [x, y, width, height]
# The function checks if keypoints fall within this bbox
ref_bbox = [10, 20, 50, 60]  # x=10, y=20, width=50, height=60

# ref_kps: list of keypoints that should be inside ref_bbox
# Each keypoint is [x, y]
ref_kps = [
    [15, 25],   # inside bbox
    [20, 30],   # inside bbox
    [35, 50],   # inside bbox
    [40, 60],   # inside bbox
    [100, 100]  # outside bbox
]

# intf_kps: list of interfering keypoints (some inside, some outside ref_bbox)
intf_kps = [
    [25, 40],   # inside bbox
    [100, 200], # outside bbox
    [30, 50]    # inside bbox
]

ARGS = {
    "ref_bbox": ref_bbox,
    "ref_kps": ref_kps,
    "intf_kps": intf_kps
}
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 float, 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
filter_bbox Ran
ran on a synthesized input
hongsukchoi/3dcrowdnet_release/tool/match_3dpw_2dpose.py
code served (permissive licence) · get_code("2a09491f993c2770")
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 dict. 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

# Create sample data matching the function's expectations
# The function expects output_list with specific structure

# Create a sample candidate pose (num_joints x 3, where last column is confidence)
candidate_pose_1 = np.array([
    [100.0, 150.0, 0.9],  # joint 0
    [110.0, 160.0, 0.8],  # joint 1
    [120.0, 170.0, 0.7],  # joint 2
    [130.0, 180.0, 0.6],  # joint 3
    [140.0, 190.0, 0.5],  # joint 4
], dtype=np.float32)

candidate_pose_2 = np.array([
    [105.0, 155.0, 0.85],
    [115.0, 165.0, 0.75],
    [125.0, 175.0, 0.65],
    [135.0, 185.0, 0.55],
    [145.0, 195.0, 0.45],
], dtype=np.float32)

# SMPL pose (target pose with valid joints)
smplpose = np.array([
    [100.0, 150.0, 0.9],
    [110.0, 160.0, 0.8],
    [120.0, 170.0, 0.7],
    [130.0, 180.0, 0.0],  # invalid joint
    [140.0, 190.0, 0.5],
], dtype=np.float32)

# OpenPose pose (reference pose)
openpose = np.array([
    [101.0, 151.0, 0.85],
    [111.0, 161.0, 0.75],
    [121.0, 171.0, 0.65],
    [131.0, 181.0, 0.0],  # invalid
    [141.0, 191.0, 0.45],
], dtype=np.float32)

# Create output list entries
output_entry_1 = {
    'candidates': [candidate_pose_1, candidate_pose_2],
    'target': {
        'openpose': openpose,
        'smplpose': smplpose,
    },
    'meta': {
        'aid': 'annotation_1',
        'img_id': 'img_001',
        'img_path': '/path/to/img1.jpg',
    }
}

# Create another entry with different candidates
candidate_pose_3 = np.array([
    [200.0, 250.0, 0.9],
    [210.0, 260.0, 0.8],
    [220.0, 270.0, 0.7],
    [230.0, 280.0, 0.6],
    [240.0, 290.0, 0.5],
], dtype=np.float32)

smplpose_2 = np.array([
    [200.0, 250.0, 0.9],
    [210.0, 260.0, 0.8],
    [220.0, 270.0, 0.0],
    [230.0, 280.0, 0.6],
    [240.0, 290.0, 0.5],
], dtype=np.float32)

openpose_2 = np.array([
    [201.0, 251.0, 0.85],
    [211.0, 261.0, 0.75],
    [221.0, 271.0, 0.0],
    [231.0, 281.0, 0.55],
    [241.0, 291.0, 0.45],
], dtype=np.float32)

output_entry_2 = {
    'candidates': [candidate_pose_3],
    'target': {
        'openpose': openpose_2,
        'smplpose': smplpose_2,
    },
    'meta': {
        'aid': 'annotation_2',
        'img_id': 'img_002',
        'img_path': '/path/to/img2.jpg',
    }
}

output_list = [output_entry_1, output_entry_2]

ARGS = {
    'output_list': output_list,
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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_bbox Ran
ran on a synthesized input
hongsukchoi/3dcrowdnet_release/tool/match_3dpw_2dpose.py
code served (permissive licence) · get_code("f8a96dbb98029759")
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 1, dtype family float, shape (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 joint_img: array of 2D image coordinates for joints
# Shape should be (num_joints, 2) where columns are [x, y]
# Using 5 joints with distinct coordinates
joint_img = np.array([
    [100.0, 150.0],
    [120.0, 180.0],
    [90.0, 200.0],
    [110.0, 160.0],
    [130.0, 190.0]
], dtype=np.float32)

# Create joint_valid: validity flags for each joint
# Shape (num_joints,) with values > 0.01 for valid joints
joint_valid = np.array([1.0, 1.0, 0.5, 1.0, 0.8], dtype=np.float32)

# extend_ratio parameter
extend_ratio = 1.2

ARGS = {
    "joint_img": joint_img,
    "joint_valid": joint_valid,
    "extend_ratio": extend_ratio
}
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 1, dtype family float, shape (4,) — shape (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
get_inclusion Ran
ran on a synthesized input
hongsukchoi/3dcrowdnet_release/tool/check_crowdidx.py
code served (permissive licence) · get_code("c10978d6702c2733")
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 int, ndim 0, dtype family int, 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
import numpy as np

torch.manual_seed(0)

# Analyzing the function:
# - bbox: indexed with [0], [1], [2], [3] - a sequence/array with 4 elements
#   bbox[0] is x coordinate, bbox[1] is y coordinate
#   bbox[2] is width, bbox[3] is height
# - kp: indexed with [0], [1] - a sequence/array with at least 2 elements
#   kp[0] is x coordinate, kp[1] is y coordinate
# The function checks if a keypoint (kp) is included in a bounding box (bbox)

# Create a bounding box [x, y, width, height]
# Using x=10, y=20, width=30, height=40
bbox = [10, 20, 30, 40]

# Create a keypoint [x, y] that is inside the bbox
# x should be between 10 and 10+30=40, y should be between 20 and 20+40=60
kp = [25, 35]

ARGS = {
    "bbox": bbox,
    "kp": kp
}
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 int, ndim 0, dtype family int, 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
make_conv1d_layers Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
hongsukchoi/3dcrowdnet_release/common/nets/layer.py
code served (permissive licence) · get_code("41843763717707cc")
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 Sequential. 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 = {
    "feat_dims": [3, 16, 32, 64],
    "kernel": 3,
    "stride": 1,
    "padding": 1,
    "bnrelu_final": 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 Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential output the recorded ndim 1 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
make_conv_layers Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
hongsukchoi/3dcrowdnet_release/common/nets/layer.py
pointer only (licence: MIT); the stored copy get_code serves, from dqj5182/contho_release, is pointer only (licence: NOASSERTION) · get_code("622140af456ff5d2")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from dqj5182/contho_release, and GitHub could not classify the licence of that repository (NOASSERTION). 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.

make_linear_layers Ran
ran; our contract draft was wrong, not the code
hongsukchoi/3dcrowdnet_release/common/nets/layer.py
pointer only (licence: MIT); the stored copy get_code serves, from dqj5182/contho_release, is pointer only (licence: NOASSERTION) · get_code("6309528f3f76d031")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from dqj5182/contho_release, and GitHub could not classify the licence of that repository (NOASSERTION). 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.

compute_iou Not yet run
unverified (harvested, no recorded run)
hongsukchoi/3dcrowdnet_release/tool/check_crowdidx.py
code served (permissive licence) · get_code("140b89f632c7ec6c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 9 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 9 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
make_conv_layers 12 other papers
1 dated earlier, 11 later
2008.03713 (2020-08)
I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image
2211.08805 (2022-11)
Interacting Hand-Object Pose Estimation via Dense Mutual Attention
and 9 more not named here
make_linear_layers 9 other papers
1 dated earlier, 8 later
2008.03713 (2020-08)
I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image
2302.00988 (2023-02)
HaMuCo: Hand Pose Estimation via Multiview Collaborative Self-Supervised Learning
and 6 more not named here
make_conv1d_layers 6 other papers
1 dated earlier, 5 later
2008.03713 (2020-08)
I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image
2302.00988 (2023-02)
HaMuCo: Hand Pose Estimation via Multiview Collaborative Self-Supervised Learning
and 3 more not named here
bbox_iou 1 other paper
1 later
Wang_3D_Human_Mesh_Recovery_with_Sequentially_Global_Rotation_Estimation_ICCV_2023_paper (2023 (year only))
3D Human Mesh Recovery with Sequentially Global Rotation Estimation
compute_CrowdIndex 1 other paper
1 later
Wang_3D_Human_Mesh_Recovery_with_Sequentially_Global_Rotation_Estimation_ICCV_2023_paper (2023 (year only))
3D Human Mesh Recovery with Sequentially Global Rotation Estimation
compute_iou 1 other paper
1 later
Wang_3D_Human_Mesh_Recovery_with_Sequentially_Global_Rotation_Estimation_ICCV_2023_paper (2023 (year only))
3D Human Mesh Recovery with Sequentially Global Rotation Estimation
filter_bbox 1 other paper
1 later
Wang_3D_Human_Mesh_Recovery_with_Sequentially_Global_Rotation_Estimation_ICCV_2023_paper (2023 (year only))
3D Human Mesh Recovery with Sequentially Global Rotation Estimation
get_bbox 1 other paper
1 later
Wang_3D_Human_Mesh_Recovery_with_Sequentially_Global_Rotation_Estimation_ICCV_2023_paper (2023 (year only))
3D Human Mesh Recovery with Sequentially Global Rotation Estimation
get_inclusion 1 other paper
1 later
Wang_3D_Human_Mesh_Recovery_with_Sequentially_Global_Rotation_Estimation_ICCV_2023_paper (2023 (year only))
3D Human Mesh Recovery with Sequentially Global Rotation Estimation

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

We consider the problem of recovering a single person's 3D human mesh from in-the-wild crowded scenes. While much progress has been in 3D human mesh estimation, existing methods struggle when test input has crowded scenes. The first reason for the failure is a domain gap between training and testing data. A motion capture dataset, which provides accurate 3D labels for training, lacks crowd data and impedes a network from learning crowded scene-robust image features of a target person. The second reason is a feature processing that spatially averages the feature map of a localized bounding box containing multiple people. Averaging the whole feature map makes a target person's feature indistinguishable from others. We present 3DCrowdNet that firstly explicitly targets in-the-wild crowded scenes and estimates a robust 3D human mesh by addressing the above issues. First, we leverage 2D human pose estimation that does not require a motion capture dataset with 3D labels for training and does not suffer from the domain gap. Second, we propose a joint-based regressor that distinguishes a target person's feature from others. Our joint-based regressor preserves the spatial activation of a target by sampling features from the target's joint locations and regresses human model parameters. As a result, 3DCrowdNet learns target-focused features and effectively excludes the irrelevant features of nearby persons. We conduct experiments on various benchmarks and prove the robustness of 3DCrowdNet to the in-the-wild crowded scenes both quantitatively and qualitatively. The code is available at https://github.com/hongsukchoi/3DCrowdNet_RELEASE.

For agents

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

get_harvested_code_for_paper("2104.07300")
get_code_for_paper("2104.07300")
have("2104.07300")

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

curl https://syntology.ai/api/ran/2104.07300.json

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

[![Syntology run record](https://syntology.ai/api/ran/2104.07300.svg)](https://syntology.ai/paper/2104.07300)

Connect an agent — have() is free.