SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.14901 · CVPR · 2022

Optimizing Elimination Templates by Greedy Parameter Search *

Tomas Pajdla, Evgeniy Martyushev, Jana Vrablikova, Spring Pajdla

arXiv · PDF · Open in the Atlas

Code that ran

8 of 8 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 8 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/2203.14901.json · badge: /api/ran/2203.14901.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
martyushev/EliminationTemplates canonical (classifier label)
classifier basis not recorded on the link
8 ran · 0 no recorded run
FunctionStatusWhere it lives
coefs_3v_triang_laurent Ran
ran on a synthesized input
martyushev/EliminationTemplates/solvers.python/py_3v_triang_laurent/coefs_3v_triang_laurent.py
code served (permissive licence) · get_code("1c7e756a82aa634a")
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 numpy as np

# The function expects `data` to be a sequence/list of 6 elements
# Each element is expected to have a .T (transpose) attribute and .flat (flattened) view
# data[0:3] are P1, P2, P3 (pattern matrices)
# data[3:6] are x1, x2, x3 (coordinate/position vectors)
# Looking at the code, elements are indexed like P1[0], P1[2], etc. up to P1[7]
# So each needs at least 8 elements when flattened
# And they're transposed, so we need 2D arrays

# Create simple 2D arrays (2x4 shape, which when transposed become 4x2, and flattened have 8 elements)
P1 = np.array([[1.0, 2.0, 3.0, 4.0],
               [5.0, 6.0, 7.0, 8.0]])

P2 = np.array([[0.5, 1.5, 2.5, 3.5],
               [4.5, 5.5, 6.5, 7.5]])

P3 = np.array([[0.1, 0.2, 0.3, 0.4],
               [0.5, 0.6, 0.7, 0.8]])

# x1, x2, x3 are accessed as x1[0], x1[1] so they need at least 2 elements
# Keep them as 1D or 2D arrays
x1 = np.array([1.0, 2.0])
x2 = np.array([3.0, 4.0])
x3 = np.array([5.0, 6.0])

# Construct data as a list of 6 elements
data = [P1, P2, P3, x1, x2, x3]

ARGS = {
    "data": data
}
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 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
coefs_5p Ran
ran on a synthesized input
martyushev/EliminationTemplates/solvers.python/py_5p/coefs_5p.py
code served (permissive licence) · get_code("fa7ee408bf223ea9")
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 numpy as np

# The function coefs_5p takes data as input
# Looking at the function body:
# - data[0] is accessed and transposed with .T
# - The result is flattened with .flat
# - Elements are accessed by index up to index 35 (A[35])
# - So A needs at least 36 elements
# - data[0].T.flat should have at least 36 elements when flattened

# Since data[0] is transposed then flattened, we need a 2D array
# If we want the flat array to have 36+ elements, we could use shape (6, 6) or similar
# Let's use a simple 6x6 array to ensure we have enough elements

data_array = np.random.randn(6, 6).astype(np.float64)

# Verify that when transposed and flattened, we get at least 36 elements
test_flat = data_array.T.flat
assert len(list(test_flat)) >= 36, f"Array too small, has {len(list(test_flat))} elements"

# Create the data structure as a list with the array as first element
data = [data_array]

ARGS = {
    "data": data
}
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 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
getT Ran
ran on a synthesized input
martyushev/EliminationTemplates/solvers.python/_common/getT.py
code served (permissive licence) · get_code("2d55ef401a2de225")
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 (4, 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

# Construct valid arguments for getT function
# M is a 2D array, RP is 1D array, B is 1D array of indices, xP is 1D array

# Create M as a matrix (rows correspond to RP values)
M = np.array([
    [1.0, 0.5, 0.2, 0.1],
    [0.3, 0.6, 0.1, 0.0],
    [0.2, 0.1, 0.9, 0.5]
])

# Create RP - labels/identifiers that we'll search in
RP = np.array([10, 20, 30])

# Create B - indices into xP (must be valid indices, so 0-3 for length 4)
B = np.array([0, 2, 1, 3])

# Create xP - array of values, some matching RP values
# xP[B[i]] will be used to look up in RP
xP = np.array([10, 5, 20, 15])

ARGS = {
    "M": M,
    "RP": RP,
    "B": B,
    "xP": xP
}
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 2, dtype family float, shape (4, 4) — shape (4, 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
ini_stats Ran
ran on a synthesized input
martyushev/EliminationTemplates/solvers.python/_common/ini_stats.py
code served (permissive licence) · get_code("4a6c9650aa1a54d6")
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 torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - ini_stats takes one parameter: prob
# - prob is used as the 'name' field in the returned dictionary
# - It appears to be a string or identifier for a problem

prob = "test_problem"

ARGS = {
    "prob": prob
}
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 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
iter_ref Ran
ran on a synthesized input
martyushev/EliminationTemplates/solvers.python/_common/iter_ref.py
code served (permissive licence) · get_code("73da5bc43ee8c870")
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 numpy as np
from scipy.linalg import lstsq, norm

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

# Define the parameters for iter_ref

# e: error vector (1D array of shape (n_modes,))
e = np.array([0.5, 0.3, 0.7, 0.2])

# S: matrix of shape (n_features, n_modes)
# This will be iteratively refined
S = np.random.randn(5, 4)

# C: matrix used in the computation (shape should allow C @ m to work)
# m will be output from U, so if m has shape (n_features,), C should be (n_output, n_features)
C = np.random.randn(3, 5)

# U: function that takes x (1D array) and returns m (1D array)
def U(x):
    # Returns a vector of shape (n_features,)
    return np.sin(x) + 0.1 * np.random.randn(5)

# dU: function that takes x (1D array) and returns dm (2D array for Jacobian)
def dU(x):
    # Returns a Jacobian matrix of shape (n_features, n_features)
    # This is used in: J = C @ dm, so dm should have shape (n_features, n_features)
    return np.cos(x)[:, np.newaxis] + 0.1 * np.random.randn(5, 5)

ARGS = {
    "e": e,
    "S": S,
    "C": C,
    "U": U,
    "dU": dU,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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
red_12x36_4pra Ran
ran on a synthesized input
martyushev/EliminationTemplates/solvers.python/py_4pra/red_12x36_4pra.py
code served (permissive licence) · get_code("4eb80f9f870da5ab")
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, 24) 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
from scipy.sparse import csr_matrix

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

# Create a simple scalar value for ss parameter
ss = 0.5

# Create C as a flat array that will be indexed by the function
# The function uses C.flat with many indices, so we need C to be large enough
# Looking at the indices used (up to 139), we need at least 140 elements total
# Create a 2D array and flatten it
C = np.random.randn(12, 36)  # 12x36 = 432 elements, more than enough

ARGS = {
    "C": C,
    "ss": ss
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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, 24) — shape (3, 24) 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
red_69x127_3v_triang_laurent Ran
ran on a synthesized input
martyushev/EliminationTemplates/solvers.python/py_3v_triang_laurent/red_69x127_3v_triang_laurent.py
code served (permissive licence) · get_code("725bb20c1e35dfb7")
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, 58) 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
from scipy.sparse import csr_matrix

# Create a small 2D numpy array to use as input
# The function indexes into C.flat, so we need an array with at least 67 elements
# (the maximum index used is 66)
C = np.random.randn(24, 3)

ARGS = {
    "C": C
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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, 58) — shape (3, 58) 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
update_stats Ran
ran on a synthesized input
martyushev/EliminationTemplates/solvers.python/_common/update_stats.py
code served (permissive licence) · get_code("7e41c8cf57f4d470")
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

# Analyze the function to understand what each parameter needs:
# - stats: a dict that needs 'err', 'flr', 'shrp', 'k', 'kr' keys (lists)
# - Ir: array that np.sum() is called on
# - e: array where e[:-1] is used, e[k] and e[k-1] are accessed, and values are compared to 0.001 and -1

# Create a small stats dict with the required keys
stats = {
    'err': [],
    'flr': [],
    'shrp': [],
    'k': [],
    'kr': []
}

# Create Ir as a small numpy array
Ir = np.array([1, 2, 3, 4])

# Create e as a numpy array
# e[:-1] will be used (all but last element)
# e[k-1] and e[k] will be accessed where k = len(e[:-1])
# If e has 5 elements, e[:-1] has 4 elements, so k=4
# We need e[3] and e[4] to be valid
# e[:-1] values should be > 0.001 for np.linalg.norm
# e[k] (e[4]) could be -1 or a positive value for the shrp calculation
e = np.array([0.1, 0.05, 0.02, 0.01, 0.005], dtype=float)

ARGS = {
    'stats': stats,
    'Ir': Ir,
    'e': e
}
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

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 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.

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 propose a new method for constructing elimination templates for efficient polynomial system solving of minimal problems in structure from motion, image matching, and camera tracking. We first construct a particular affine parameterization of the elimination templates for systems with a finite number of distinct solutions. Then, we use a heuristic greedy optimization strategy over the space of parameters to get a template with a small size. We test our method on 34 minimal problems in computer vision. For all of them, we found the templates either of the same or smaller size compared to the state-of-the-art. For some difficult examples, our templates are, e.g., 2.1, 2.5, 3.8, 6.6 times smaller. For the problem of refractive absolute pose estimation with unknown focal length, we have found a template that is 20 times smaller. Our experiments on synthetic data also show that the new solvers are fast and numerically accurate. We also present a fast and numerically accurate solver for the problem of relative pose estimation with unknown common focal length and radial distortion.

For agents

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

get_harvested_code_for_paper("2203.14901")
get_code_for_paper("2203.14901")
have("2203.14901")

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

curl https://syntology.ai/api/ran/2203.14901.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.14901.svg)](https://syntology.ai/paper/2203.14901)

Connect an agent — have() is free.