SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2302.11566 · CVPR · 2023

Vid2Avatar: 3D Avatar Reconstruction from Videos in the Wild via Self-supervised Scene Decomposition

Otmar Hilliges, Xu Chen, Jie Song, Chen Guo, Tianjian Jiang

arXiv · PDF · Open in the Atlas

Code that ran

7 of 7 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; 1 honoured the contract we drafted, 0 violated it and 6 have no contract verdict; 0 failures of our instrument; 0 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/2302.11566.json · badge: /api/ran/2302.11566.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
MoyGcc/vid2avatar canonical (classifier label)
classifier basis not recorded on the link
7 ran · 0 no recorded run
FunctionStatusWhere it lives
estimate_translation_cv2 Ran
ran on a synthesized input
MoyGcc/vid2avatar/preprocessing/preprocessing_utils.py
code served (permissive licence) · get_code("66473bc0db1ad9b6")
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 (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
import cv2

# Set up small valid test data for the function
# joints_3d: 3D joint coordinates (N x 3), should be in camera space
# joints_2d: 2D image coordinates (N x 2)
# We need at least 4 points for solvePnPRansac

# Create 5 sample 3D points
joints_3d = np.array([
    [0.0, 0.0, 5.0],
    [0.1, 0.0, 5.0],
    [-0.1, 0.0, 5.0],
    [0.0, 0.1, 5.0],
    [0.0, -0.1, 5.0]
], dtype=np.float32)

# Create corresponding 2D image projections
# For a point (x, y, z) with focal_length f and principal point (cx, cy):
# u = f * x / z + cx
# v = f * y / z + cy
focal_length = 600.0
img_size = np.array([512.0, 512.0])
principal_point = img_size // 2  # [256, 256]

joints_2d = np.array([
    [256.0, 256.0],           # (0, 0, 5) projects to center
    [268.0, 256.0],           # (0.1, 0, 5) with focal_length 600
    [244.0, 256.0],           # (-0.1, 0, 5)
    [256.0, 268.0],           # (0, 0.1, 5)
    [256.0, 244.0]            # (0, -0.1, 5)
], dtype=np.float32)

# Build projection matrix
camK = np.eye(3, dtype=np.float32)
camK[0, 0] = focal_length
camK[1, 1] = focal_length
camK[0, 2] = principal_point[0]
camK[1, 2] = principal_point[1]

ARGS = {
    "joints_3d": joints_3d,
    "joints_2d": joints_2d,
    "focal_length": 600,
    "img_size": img_size,
    "proj_mat": None,
    "cam_dist": None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 (3,) — shape (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_bbox_center Ran
ran on a synthesized input
MoyGcc/vid2avatar/preprocessing/run_openpose.py
code served (permissive licence) · get_code("2cfa68461442724d")
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 (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 cv2
import numpy as np
import tempfile
import os

# Create temporary directory for test files
temp_dir = tempfile.mkdtemp()

# Create a simple test image (100x80 RGB image)
img = np.ones((80, 100, 3), dtype=np.uint8) * 200
img_path = os.path.join(temp_dir, "test_img.png")
cv2.imwrite(img_path, img)

# Create a simple mask image with a white region in the center
mask = np.zeros((80, 100, 3), dtype=np.uint8)
mask[20:60, 30:70, :] = 255  # White region from (30,20) to (70,60)
mask_path = os.path.join(temp_dir, "test_mask.png")
cv2.imwrite(mask_path, mask)

ARGS = {
    "img_path": img_path,
    "mask_path": mask_path,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 (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
get_center_point Ran
ran on a synthesized input
MoyGcc/vid2avatar/preprocessing/normalize_cameras.py
code served (permissive licence) · get_code("ef76c0d2137da237")
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 (3, 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 cv2
import numpy as np

# Create projection matrices for cameras
# A projection matrix P is typically 3x4
num_cams = 2

# Create valid projection matrices
# These should be 3x4 matrices that cv2.decomposeProjectionMatrix can handle
K_camera = np.array([
    [500, 0, 320],
    [0, 500, 240],
    [0, 0, 1]
], dtype=np.float32)

R_camera = np.eye(3, dtype=np.float32)

t_camera_1 = np.array([[0], [0], [0]], dtype=np.float32)
t_camera_2 = np.array([[1], [0], [0]], dtype=np.float32)

P1 = K_camera @ np.hstack([R_camera, t_camera_1])
P2 = K_camera @ np.hstack([R_camera, t_camera_2])

cameras = {
    'cam_0': P1,
    'cam_1': P2
}

ARGS = {
    'num_cams': num_cams,
    'cameras': cameras
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 (3, 2) — shape (3, 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
pose_temporal_loss Ran
ran on a synthesized input
MoyGcc/vid2avatar/preprocessing/loss.py
code served (permissive licence) · get_code("5a5d56a98920d8db")
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)

# The function takes two pose tensors and computes mean squared difference
# Both parameters should be tensors of the same shape
last_pose = torch.randn(2, 4)
param_pose = torch.randn(2, 4)

ARGS = {
    "last_pose": last_pose,
    "param_pose": param_pose,
}
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
render_trimesh Ran
ran on a synthesized input
MoyGcc/vid2avatar/preprocessing/preprocessing_utils.py
code served (permissive licence) · get_code("dfdc897db8cbf77c")
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 int, shape (256, 256, 3) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
from unittest.mock import Mock

torch.manual_seed(0)

# Create a mock mesh object with required attributes
mock_mesh = Mock()
mock_mesh.vertices = np.random.randn(10, 3).astype(np.float32)
mock_mesh.faces = np.array([[0, 1, 2], [1, 2, 3], [2, 3, 4]], dtype=np.int32)
mock_mesh.visual.vertex_colors = np.random.randint(0, 256, (10, 4), dtype=np.uint8)

# Create a mock renderer with required methods
mock_renderer = Mock()
mock_renderer.set_camera = Mock()
mock_renderer.render_mesh_recon = Mock(
    return_value=torch.rand(1, 256, 256, 3)
)

# Create rotation and translation matrices
R = np.eye(3, dtype=np.float32)
T = np.zeros(3, dtype=np.float32)

ARGS = {
    'renderer': mock_renderer,
    'mesh': mock_mesh,
    'R': R,
    'T': T,
    'mode': 'np'
}
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 ndarray, ndim 3, dtype family int, shape (256, 256, 3) — shape (256, 256, 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
skinning Ran
ran on a synthesized input
MoyGcc/vid2avatar/code/lib/model/deformer.py
code served (permissive licence) · get_code("9e85c5e2bf9ccd61")
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, 4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn.functional as F

torch.manual_seed(0)

# Define dimensions
B = 2  # batch size
N = 4  # number of points
D = 3  # spatial dimension
J = 5  # number of joints

# Create canonical points [B, N, D]
x = torch.randn(B, N, D)

# Create weights [B, N, J] - blend weights for each point across joints
w = torch.softmax(torch.randn(B, N, J), dim=-1)

# Create bone transformation matrices [B, J, D+1, D+1]
# These are 4x4 homogeneous transformation matrices
tfs = torch.randn(B, J, D+1, D+1)

# Make transformation matrices invertible by using random orthogonal-like matrices
for b in range(B):
    for j in range(J):
        # Create a simple invertible matrix
        tfs[b, j] = torch.eye(D+1) + 0.1 * torch.randn(D+1, D+1)

# Test with inverse=False
inverse_flag = False

ARGS = {
    'x': x,
    'w': w,
    'tfs': tfs,
    'inverse': inverse_flag
}
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, 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
smpl_to_pose Ran
ran, honoured the contract we drafted
MoyGcc/vid2avatar/preprocessing/preprocessing_utils.py
code served (permissive licence) · get_code("d5dd5e6e08a6dc7c")
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 int, shape (118,) 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

ARGS = {
    'model_type': 'smplx',
    'use_hands': True,
    'use_face': True,
    'use_face_contour': False,
    'openpose_format': 'coco25',
}
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 int, shape (118,) — shape (118,) 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

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 6 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
smpl_to_pose 1 other paper
1 dated earlier
2111.14672 (2021-11)
Human Performance Capture from Monocular Video in the Wild

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

Figure 1. We propose Vid2Avatar, a method to reconstruct detailed 3D avatars from monocular videos in the wild via self-supervised scene decomposition. Our method does not require any groundtruth supervision or priors extracted from large datasets of clothed human scans, nor do we rely on any external segmentation modules. Project page: https://ait.ethz.ch/vid2avatar.

For agents

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

get_harvested_code_for_paper("2302.11566")
get_code_for_paper("2302.11566")
have("2302.11566")

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

curl https://syntology.ai/api/ran/2302.11566.json

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

[![Syntology run record](https://syntology.ai/api/ran/2302.11566.svg)](https://syntology.ai/paper/2302.11566)

Connect an agent — have() is free.