SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.08367 · ICCV · 2021

SO-Pose: Exploiting Self-Occlusion for Direct 6D Pose Estimation

Federico Tombari, Nassir Navab, Xiangyang Ji, Yan Di, Fabian Manhardt, Gu Wang

arXiv · PDF · Open in the Atlas

Code that ran

13 of 16 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 10 have no contract verdict; 3 failures of our instrument; 3 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/2108.08367.json · badge: /api/ran/2108.08367.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
shangbuhuan13/so-pose canonical (classifier label)
classifier basis: census
5 ran · 1 no recorded run
THU-DA-6D-Pose-Group/GDR-Net community (classifier label)
classifier basis not recorded on the link
8 ran · 2 no recorded run
FunctionStatusWhere it lives
SoftPool Ran
ran on a synthesized input
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/models/point_pnp_net.py
code served (permissive licence) · get_code("dee3e76e2183b408")
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, 4, 2, 4) 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 torch.nn.functional as F

torch.manual_seed(0)

# The function expects:
# x: (B, F, P) tensor
# N_p: int (top k parameter)

# Create a small test tensor
# B=2 (batch size)
# F=4 (feature dimension)
# P=3 (number of points/positions)
x = torch.randn(2, 4, 3)

# N_p should be an integer <= P
N_p = 2

ARGS = {
    "x": x,
    "N_p": N_p,
}
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, 4, 2, 4) — shape (2, 4, 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
batch_data_test Ran
ran on a synthesized input
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/engine_utils.py
code served (permissive licence) · get_code("b048eee7d945442d")
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. 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 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 itertools

torch.manual_seed(0)

# Create a minimal config dict
cfg = {
    'model': 'test_model',
    'batch_size': 2,
}

# Create sample data items
data_item_1 = {
    'im_H': torch.tensor([[480.0]], dtype=torch.float32),
    'im_W': torch.tensor([[640.0]], dtype=torch.float32),
    'roi_img': torch.randn(1, 3, 32, 32, dtype=torch.float32),
    'inst_id': torch.tensor([[1]], dtype=torch.long),
    'roi_coord_2d': torch.randn(1, 4, dtype=torch.float32),
    'roi_cls': torch.tensor([[2]], dtype=torch.long),
    'score': torch.tensor([[0.9]], dtype=torch.float32),
    'roi_extent': torch.randn(1, 3, dtype=torch.float32),
    'bbox': torch.randn(1, 4, dtype=torch.float32),
    'bbox_est': torch.randn(1, 4, dtype=torch.float32),
    'bbox_mode': torch.tensor([[0]], dtype=torch.float32),
    'roi_wh': torch.randn(1, 2, dtype=torch.float32),
    'scale': torch.tensor([[1.0]], dtype=torch.float32),
    'resize_ratio': torch.tensor([[1.0]], dtype=torch.float32),
    'cam': torch.randn(1, 4, 4, dtype=torch.float32),
    'bbox_center': torch.randn(1, 3, dtype=torch.float32),
    'scene_im_id': ['scene_001'],
    'file_name': ['image_001.jpg'],
    'model_info': ['model_a'],
}

data_item_2 = {
    'im_H': torch.tensor([[480.0]], dtype=torch.float32),
    'im_W': torch.tensor([[640.0]], dtype=torch.float32),
    'roi_img': torch.randn(1, 3, 32, 32, dtype=torch.float32),
    'inst_id': torch.tensor([[2]], dtype=torch.long),
    'roi_coord_2d': torch.randn(1, 4, dtype=torch.float32),
    'roi_cls': torch.tensor([[3]], dtype=torch.long),
    'score': torch.tensor([[0.85]], dtype=torch.float32),
    'roi_extent': torch.randn(1, 3, dtype=torch.float32),
    'bbox': torch.randn(1, 4, dtype=torch.float32),
    'bbox_est': torch.randn(1, 4, dtype=torch.float32),
    'bbox_mode': torch.tensor([[1]], dtype=torch.float32),
    'roi_wh': torch.randn(1, 2, dtype=torch.float32),
    'scale': torch.tensor([[1.5]], dtype=torch.float32),
    'resize_ratio': torch.tensor([[0.5]], dtype=torch.float32),
    'cam': torch.randn(1, 4, 4, dtype=torch.float32),
    'bbox_center': torch.randn(1, 3, dtype=torch.float32),
    'scene_im_id': ['scene_002'],
    'file_name': ['image_002.jpg'],
    'model_info': ['model_b'],
}

data = [data_item_1, data_item_2]

ARGS = {
    'cfg': cfg,
    'data': data,
    'device': 'cpu',
}
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 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
conv3x3 Ran
ran on a synthesized input
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/models/pvnet_net/resnet.py
code served (permissive licence) · get_code("861bfac8e714414f")
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 Conv2d. 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 numpy as np

torch.manual_seed(0)

ARGS = {
    'in_planes': 3,
    'out_planes': 64,
    'stride': 1,
    'dilation': 1
}
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 Conv2d — a Conv2d output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Conv2d 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
cou Ran
ran on a synthesized input
shangbuhuan13/so-pose/lib/pysixd/pose_error_more.py
code served (permissive licence) · get_code("787ea5208581b998")
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
import torch
import torch.nn as nn

torch.manual_seed(0)

# Create a minimal renderer class that implements the render method
class MinimalRenderer:
    def render(self, obj_id, R, t, K=None, W=None, H=None, near=None, far=None, to_255=False):
        """Minimal render method that returns RGB and depth images."""
        # Return dummy RGB (not used) and depth image
        depth = np.random.rand(H, W).astype(np.float32) * 100
        # Make some pixels valid (> 0) and some invalid (= 0)
        depth[5:15, 5:15] = np.random.rand(10, 10) * 100 + 1
        rgb = np.zeros((H, W, 3), dtype=np.uint8)
        return rgb, depth

# Create rotation matrix (3x3)
R_est = np.array([
    [0.9, -0.436, 0.0],
    [0.436, 0.9, 0.0],
    [0.0, 0.0, 1.0]
], dtype=np.float32)

# Create translation vector (3x1)
t_est = np.array([[0.1], [0.2], [5.0]], dtype=np.float32)

# Ground truth pose
R_gt = np.array([
    [1.0, 0.0, 0.0],
    [0.0, 1.0, 0.0],
    [0.0, 0.0, 1.0]
], dtype=np.float32)

t_gt = np.array([[0.0], [0.0], [5.0]], dtype=np.float32)

# Create model dict with 3D points
model = {
    'pts': np.random.rand(10, 3).astype(np.float32)
}

# Image size (width, height)
im_size = (640, 480)

# Camera intrinsic matrix
K = np.array([
    [500.0, 0.0, 320.0],
    [0.0, 500.0, 240.0],
    [0.0, 0.0, 1.0]
], dtype=np.float32)

# Renderer instance
renderer = MinimalRenderer()

ARGS = {
    'R_est': R_est,
    't_est': t_est,
    'R_gt': R_gt,
    't_gt': t_gt,
    'model': model,
    'im_size': im_size,
    'K': K,
    'renderer': renderer
}
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 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_out_coor Ran
ran on a synthesized input
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/engine_utils.py
code served (permissive licence) · get_code("2aae5b4a58333958")
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 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 torch

torch.manual_seed(0)

# Build a minimal config object that has the required attributes
class MinimalConfig:
    class MODEL:
        class CDPN:
            class ROT_HEAD:
                XYZ_BIN = 8
    
cfg = MinimalConfig()

# Test case 1: single values per batch (shape[1] == 1)
batch_size = 2
coor_x = torch.randn(batch_size, 1)
coor_y = torch.randn(batch_size, 1)
coor_z = torch.randn(batch_size, 1)

ARGS = {
    "cfg": cfg,
    "coor_x": coor_x,
    "coor_y": coor_y,
    "coor_z": coor_z,
}
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 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
match_poses_scene Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
shangbuhuan13/so-pose/lib/pysixd/pose_matching.py
code served (permissive licence) · get_code("e396a9fc5071ee61")
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 1, 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

# Define ARGS for match_poses_scene
ARGS = {
    'scene_id': 0,
    'scene_gt': {
        0: [
            {'obj_id': 1},
            {'obj_id': 2},
        ],
        1: [
            {'obj_id': 1},
        ],
    },
    'scene_gt_valid': {
        0: [True, True],
        1: [True],
    },
    'scene_errs': [
        {
            'im_id': 0,
            'obj_id': 1,
            'est_id': 10,
            'score': 0.9,
            'errors': {0: [0.5, 0.3]},
        },
        {
            'im_id': 0,
            'obj_id': 1,
            'est_id': 11,
            'score': 0.7,
            'errors': {0: [0.8, 0.6]},
        },
        {
            'im_id': 0,
            'obj_id': 2,
            'est_id': 20,
            'score': 0.85,
            'errors': {1: [0.4, 0.2]},
        },
        {
            'im_id': 1,
            'obj_id': 1,
            'est_id': 30,
            'score': 0.95,
            'errors': {0: [0.3, 0.2]},
        },
    ],
    'correct_th': [1.0, 1.0],
    'n_top': 2,
}
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 list, ndim 1, dtype family float — 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
qmul_torch Ran
ran on a synthesized input
shangbuhuan13/so-pose/core/utils/pose_utils.py
code served (permissive licence) · get_code("08d2425b782252f2")
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 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 torch

torch.manual_seed(0)

# Create two quaternion tensors of shape (2, 4)
# Each row is a quaternion [w, x, y, z]
q = torch.tensor([
    [1.0, 0.0, 0.0, 0.0],
    [0.7071, 0.7071, 0.0, 0.0]
], dtype=torch.float32)

r = torch.tensor([
    [1.0, 0.0, 0.0, 0.0],
    [0.7071, 0.0, 0.7071, 0.0]
], dtype=torch.float32)

ARGS = {
    "q": q,
    "r": r
}
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 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
qrot_points_th Ran
ran on a synthesized input
shangbuhuan13/so-pose/core/utils/pose_utils.py
code served (permissive licence) · get_code("51f6831774e2dafc")
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 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 torch

torch.manual_seed(0)

# q: (4,) - a quaternion
q = torch.tensor([1.0, 0.1, 0.2, 0.3])

# points: (N, 3) - N points in 3D space
points = torch.tensor([
    [1.0, 0.0, 0.0],
    [0.0, 1.0, 0.0]
])

ARGS = {
    "q": q,
    "points": 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 Tensor, 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
qrot_torch Ran
ran on a synthesized input
shangbuhuan13/so-pose/core/utils/pose_utils.py
code served (permissive licence) · get_code("392731415f67f130")
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 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 torch

torch.manual_seed(0)

# Create quaternion tensor of shape (2, 4)
# Quaternions are represented as [w, x, y, z] where w is the scalar part
q = torch.tensor([
    [1.0, 0.0, 0.0, 0.0],  # identity quaternion
    [0.9239, 0.3827, 0.0, 0.0]  # rotation around z-axis
], dtype=torch.float32)

# Create vector tensor of shape (2, 3)
v = torch.tensor([
    [1.0, 0.0, 0.0],
    [0.0, 1.0, 0.0]
], dtype=torch.float32)

ARGS = {
    "q": q,
    "v": v
}
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 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
resnet18 Ran
ran; our contract draft was wrong, not the code
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/models/pvnet_net/resnet.py
code served (permissive licence) · get_code("6fae332aec4c1b22")
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 ResNet. 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 math
import torch.utils.model_zoo as model_zoo
import numpy as np

torch.manual_seed(0)

# Define ARGS dict with all parameters for resnet18
ARGS = {
    'pretrained': False,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
resnet34 Ran
ran; our contract draft was wrong, not the code
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/models/pvnet_net/resnet.py
code served (permissive licence) · get_code("f8f084c1303e383e")
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 ResNet. 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 math
import numpy as np

torch.manual_seed(0)

# The function resnet34 has the following signature:
# def resnet34(pretrained=False, **kwargs):
# 
# Parameters:
# - pretrained: bool, defaults to False
# - **kwargs: additional keyword arguments passed to ResNet.__init__
#
# ResNet.__init__ parameters are:
# - block: the block class (BasicBlock)
# - layers: list of layer counts [3, 4, 6, 3]
# - num_classes: int, defaults to 1000
# - fully_conv: bool, defaults to False
# - remove_avg_pool_layer: bool, defaults to False
# - output_stride: int, defaults to 32

ARGS = {
    'pretrained': False,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
topk_pool Ran
ran on a synthesized input
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/models/conv_pnp_net.py
code served (permissive licence) · get_code("c68cad7f699e4a8d")
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, 4, 3, 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)

# x should be shape (B, F, M, P)
# B=2 (batch), F=4 (features), M=3 (some dimension), P=5 (pool dimension)
x = torch.randn(2, 4, 3, 5)

# k should be an integer <= P (the last dimension size)
k = 3

ARGS = {
    "x": x,
    "k": k,
}
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, 4, 3, 3) — shape (2, 4, 3, 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
topk_pool_with_region Ran
ran on a synthesized input
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/models/conv_pnp_net.py
code served (permissive licence) · get_code("545579359398958d")
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, 4, 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 torch

torch.manual_seed(0)

# x: (B, F, M, P) - batch size 2, features 4, M=3, P=5
x = torch.randn(2, 4, 3, 5)

# region: (B, M, P) - batch size 2, M=3, P=5
# This tensor will be used with torch.topk, so it should have numeric values
region = torch.randn(2, 3, 5)

# k: top k value
k = 2

ARGS = {
    "x": x,
    "region": region,
    "k": k,
}
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, 4, 3, 2) — shape (2, 4, 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
batch_data Not yet run
unverified (harvested, no recorded run)
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/engine_utils.py
code served (permissive licence) · get_code("7d3d30f20b202337")
How it ran

no invocation recorded.

get_available_datasets Not yet run
unverified (harvested, no recorded run)
THU-DA-6D-Pose-Group/GDR-Net/core/gdrn_modeling/dataset_factory.py
code served (permissive licence) · get_code("91327f54d1b07174")
How it ran

no invocation recorded.

match_poses Not yet run
unverified (harvested, no recorded run)
shangbuhuan13/so-pose/lib/pysixd/pose_matching.py
code served (permissive licence) · get_code("b441bb9b0ad6e860")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 4 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 2 the same month, 4 later) and 12 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
resnet18 89 other papers
35 dated earlier, 1 the same month, 51 later, 2 same year, order not determinable
1705.09759 (2017-05)
CASENet: Deep Category-Aware Semantic Edge Detection
1801.07698 (2018-01)
ArcFace: Additive Angular Margin Loss for Deep Face Recognition
1803.00839 (2018-03)
Pose-Robust Face Recognition via Deep Residual Equivariant Mapping
and 86 more not named here
resnet34 53 other papers
28 dated earlier, 1 the same month, 22 later, 2 same year, order not determinable
1705.09759 (2017-05)
CASENet: Deep Category-Aware Semantic Edge Detection
1801.07698 (2018-01)
ArcFace: Additive Angular Margin Loss for Deep Face Recognition
1803.00839 (2018-03)
Pose-Robust Face Recognition via Deep Residual Equivariant Mapping
and 50 more not named here
match_poses 2 other papers
2 later
2505.04109 (2025-05)
One2Any: One-Reference 6D Pose Estimation for Any Object
2601.13565 (2026-01)
Learning Fine-Grained Correspondence with Cross-Perspective Perception for Open-Vocabulary 6D Object Pose Estimation
match_poses_scene 2 other papers
2 later
2505.04109 (2025-05)
One2Any: One-Reference 6D Pose Estimation for Any Object
2601.13565 (2026-01)
Learning Fine-Grained Correspondence with Cross-Perspective Perception for Open-Vocabulary 6D Object Pose 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

Directly regressing all 6 degrees-of-freedom (6DoF) for the object pose (i.e. the 3D rotation and translation) in a cluttered environment from a single RGB image is a challenging problem. While end-to-end methods have recently demonstrated promising results at high efficiency, they are still inferior when compared with elaborate PnP/RANSACbased approaches in terms of pose accuracy. In this work, we address this shortcoming by means of a novel reasoning about self-occlusion, in order to establish a two-layer representation for 3D objects which considerably enhances the accuracy of end-to-end 6D pose estimation. Our framework, named SO-Pose, takes a single RGB image as input and respectively generates 2D-3D correspondences as well as self-occlusion information harnessing a shared encoder and two separate decoders. Both outputs are then fused to directly regress the 6DoF pose parameters. Incorporating cross-layer consistencies that align correspondences, selfocclusion and 6D pose, we can further improve accuracy and robustness, surpassing or rivaling all other state-ofthe-art approaches on various challenging datasets.

For agents

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

get_harvested_code_for_paper("2108.08367")
get_code_for_paper("2108.08367")
have("2108.08367")

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

curl https://syntology.ai/api/ran/2108.08367.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.08367.svg)](https://syntology.ai/paper/2108.08367)

Connect an agent — have() is free.