SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.00633 · ICCV · 2021

RePOSE: Fast 6D Object Pose Refinement via Deep Texture Rendering

Kris Kitani, Xingyu Liu, Shun Iwase, Rio Yokota, Rawal Khirodkar

arXiv · PDF · Open in the Atlas

Code that ran

8 of 10 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 4 have no contract verdict; 4 failures of our instrument; 2 no recorded run), measured 2026-11-22; 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/2104.00633.json · badge: /api/ran/2104.00633.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
sh8/repose canonical (classifier label)
classifier basis: census
8 ran · 2 no recorded run
FunctionStatusWhere it lives
make_batch_data_sampler Ran
ran on a synthesized input
sh8/repose/lib/datasets/make_dataset.py
code served (permissive licence) · get_code("872ab7704b4526e2")
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 BatchSampler. 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.utils.data
from torch.utils.data.sampler import Sampler

torch.manual_seed(0)

# Create a minimal sampler
class DummySampler(Sampler):
    def __init__(self, size=10):
        self.size = size
    
    def __iter__(self):
        return iter(range(self.size))
    
    def __len__(self):
        return self.size

# Create a minimal samplers module with required classes
class DummySamplers:
    class IterationBasedBatchSampler:
        def __init__(self, batch_sampler, max_iter):
            self.batch_sampler = batch_sampler
            self.max_iter = max_iter
    
    class ImageSizeBatchSampler:
        def __init__(self, sampler, batch_size, drop_last, *args):
            self.sampler = sampler
            self.batch_size = batch_size
            self.drop_last = drop_last
            self.args = args

samplers = DummySamplers()

# Create a minimal config object
class Config:
    class Train:
        sampler = 'default'
    train = Train()

cfg = Config()

# Create ARGS dictionary
ARGS = {
    'cfg': cfg,
    'sampler': DummySampler(size=10),
    'batch_size': 2,
    'drop_last': False,
    'max_iter': -1,
    'is_train': 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 BatchSampler — a BatchSampler output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a BatchSampler 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
make_collator Ran
ran; our contract draft was wrong, not the code
sh8/repose/lib/datasets/collate_batch.py
code served (permissive licence) · get_code("a11a27a9c4153d8d")
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 function. 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)

class MockConfig:
    def __init__(self, task):
        self.task = task

cfg = MockConfig(task="test_task")

ARGS = {
    "cfg": cfg
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
make_data_sampler Ran
ran on a synthesized input
sh8/repose/lib/datasets/make_dataset.py
code served (permissive licence) · get_code("eb9f81ec5588ee75")
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 RandomSampler. 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.utils.data

torch.manual_seed(0)

# Create a simple dataset
class SimpleDataset(torch.utils.data.Dataset):
    def __init__(self, size=10):
        self.size = size
    
    def __len__(self):
        return self.size
    
    def __getitem__(self, idx):
        return torch.randn(4), torch.randint(0, 2, (1,))

dataset = SimpleDataset(size=10)

ARGS = {
    "dataset": dataset,
    "shuffle": 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 RandomSampler — a RandomSampler output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a RandomSampler 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
make_network Ran
ran on a synthesized input
sh8/repose/lib/networks/make_network.py
code served (permissive licence) · get_code("e941a1cab31db49e")
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 function. 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 os
import sys
import types

# Create a minimal config object with a 'task' attribute
class Config:
    def __init__(self, task):
        self.task = task

# Create a mock module structure to avoid file I/O
# We'll create a fake network module in memory
mock_network_module = types.ModuleType('lib.networks.test_task')

def mock_get_network(cfg):
    """Mock get_network function that returns a simple callable"""
    return lambda x: x

mock_network_module.get_network = mock_get_network

# Monkey-patch sys.modules to provide the fake module
sys.modules['lib'] = types.ModuleType('lib')
sys.modules['lib.networks'] = types.ModuleType('lib.networks')
sys.modules['lib.networks.test_task'] = mock_network_module

# Create a temporary directory structure and files for imp.load_source to work
# Actually, we need to make imp.load_source work without real files
# Let's override imp.load_source to return our mock module
import imp as imp_module

original_load_source = imp_module.load_source

def mock_load_source(module_name, pathname):
    """Mock load_source that returns our fake module"""
    return mock_network_module

imp_module.load_source = mock_load_source

# Now construct the config for the function
cfg = Config(task='test_task')

ARGS = {
    'cfg': cfg
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
imp — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
make_transforms Ran
ran on a synthesized input
sh8/repose/lib/datasets/transforms.py
code served (permissive licence) · get_code("2d5b772e8d9d2dcf")
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 Compose. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import random
from torchvision import transforms
import cv2

class RandomBlur(object):
    def __init__(self, prob=0.5):
        self.prob = prob

    def __call__(self, image):
        if random.random() < self.prob:
            sigma = np.random.choice([3, 5, 7, 9])
            image = cv2.GaussianBlur(image, (sigma, sigma), 0)
        return image

def make_transforms(cfg, is_train):
    if is_train is True:
        transform = transforms.Compose([RandomBlur(0.5), transforms.ToTensor(), transforms.ColorJitter(0.1, 0.1, 0.05, 0.05), transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
    else:
        transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
    return transform

cfg = {
    "model": "resnet50",
    "num_classes": 10,
    "learning_rate": 0.001
}

ARGS = {
    "cfg": cfg,
    "is_train": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Compose — a Compose output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Compose 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
resize_keep_aspect_ratio Ran
ran; our contract draft was wrong, not the code
sh8/repose/lib/datasets/augmentation.py
code served (permissive licence) · get_code("469f96dc37b6cbbf")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import cv2
import numpy as np

# Create a small test image with distinct dimensions
# Using 5x4 to have distinguishable height and width
img = np.random.randint(0, 256, (5, 4, 3), dtype=np.uint8)

# Define arguments for the function
ARGS = {
    'img': img,
    'imsize': 8,
    'intp_type': cv2.INTER_LINEAR
}
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 tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rotate Ran
ran; our contract draft was wrong, not the code
sh8/repose/lib/datasets/augmentation.py
code served (permissive licence) · get_code("317fd27af8e8a5f6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import cv2
import numpy as np

# Create a small test image (height=100, width=150, channels=3)
img = np.random.randint(0, 256, (100, 150, 3), dtype=np.uint8)

# Create a small test mask (same dimensions as img, single channel)
mask = np.random.randint(0, 2, (100, 150), dtype=np.uint8)

# Create homography coordinates - typically a 4x3 matrix of float32
# representing transformed points in homogeneous coordinates
hcoords = np.array([
    [1.0, 0.0, 0.0],
    [0.0, 1.0, 0.0],
    [0.0, 0.0, 1.0],
    [0.5, 0.5, 1.0]
], dtype=np.float32)

# Rotation angle range in degrees
rot_ang_min = -15.0
rot_ang_max = 15.0

ARGS = {
    "img": img,
    "mask": mask,
    "hcoords": hcoords,
    "rot_ang_min": rot_ang_min,
    "rot_ang_max": rot_ang_max
}
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 tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rotate_instance Ran
ran; our contract draft was wrong, not the code
sh8/repose/lib/datasets/augmentation.py
code served (permissive licence) · get_code("8ac07ff4bce4b058")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import cv2
import numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create a small test image (64x48 RGB)
img = np.random.randint(0, 256, (64, 48, 3), dtype=np.uint8)

# Create a small mask with some nonzero values
mask = np.zeros((64, 48), dtype=np.uint8)
mask[20:40, 15:35] = 1  # Rectangle region in the mask

# Create a 3x3 rotation matrix (identity)
rot = np.eye(3, dtype=np.float32)

# Create another 3x3 rotation matrix for initial rotation
rot_ini = np.eye(3, dtype=np.float32)

# Rotation angle parameters
rot_ang_min = -15.0
rot_ang_max = 15.0

# rot parameter (unused in function but part of signature)
rot = True

ARGS = {
    "img": img,
    "mask": mask,
    "rot": np.eye(3, dtype=np.float32),
    "rot_ini": np.eye(3, dtype=np.float32),
    "rot_ang_min": rot_ang_min,
    "rot_ang_max": rot_ang_max,
}
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 tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
make_analyzer Not yet run
unverified (harvested, no recorded run)
sh8/repose/lib/analyzers/make_analyzer.py
code served (permissive licence) · get_code("e2a61f3fba0c0ec3")
How it ran

no invocation recorded.

make_dataset Not yet run
unverified (harvested, no recorded run)
sh8/repose/lib/datasets/make_dataset.py
code served (permissive licence) · get_code("9cc25f2c8373bccc")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 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
make_analyzer 1 other paper
1 dated earlier
1812.11788 (2018-12)
PVNet: Pixel-wise Voting Network for 6DoF Pose Estimation
make_collator 1 other paper
1 dated earlier
1812.11788 (2018-12)
PVNet: Pixel-wise Voting Network for 6DoF Pose Estimation
resize_keep_aspect_ratio 1 other paper
1 dated earlier
1812.11788 (2018-12)
PVNet: Pixel-wise Voting Network for 6DoF Pose Estimation
rotate 1 other paper
1 dated earlier
1812.11788 (2018-12)
PVNet: Pixel-wise Voting Network for 6DoF 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

We present RePOSE, a fast iterative refinement method for 6D object pose estimation. Prior methods perform refinement by feeding zoomed-in input and rendered RGB images into a CNN and directly regressing an update of a refined pose. Their runtime is slow due to the computational cost of CNN, which is especially prominent in multiple-object pose refinement. To overcome this problem, RePOSE leverages image rendering for fast feature extraction using a 3D model with a learnable texture. We call this deep texture rendering, which uses a shallow multilayer perceptron to directly regress a view-invariant image representation of an object. Furthermore, we utilize differentiable Levenberg-Marquardt (LM) optimization to refine a pose fast and accurately by minimizing the distance between the input and rendered image representations without the need of zooming in. These image representations are trained such that differentiable LM optimization converges within few iterations. Consequently, RePOSE runs at 92 FPS and achieves state-of-the-art accuracy of 51.6% on the Occlusion LineMOD dataset -a 4.1% absolute improvement over the prior art, and comparable result on the YCB-Video dataset with a much faster runtime. The code is available at https://github.com/sh8/repose.

For agents

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

get_harvested_code_for_paper("2104.00633")
get_code_for_paper("2104.00633")
have("2104.00633")

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

curl https://syntology.ai/api/ran/2104.00633.json

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

[![Syntology run record](https://syntology.ai/api/ran/2104.00633.svg)](https://syntology.ai/paper/2104.00633)

Connect an agent — have() is free.