SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.02733 · 2021

DISCO: accurate Discrete Scale Convolutions

arXiv · PDF · Open in the Atlas

Code that ran

11 of 17 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 11 have no contract verdict; 0 failures of our instrument; 6 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/2106.02733.json · badge: /api/ran/2106.02733.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
isosnovik/disco canonical (classifier label)
classifier basis: census
11 ran · 6 no recorded run
FunctionStatusWhere it lives
apply_rescaling Ran
ran on a synthesized input
isosnovik/disco/models/basis/disco.py
code served (permissive licence) · get_code("4c9cb0752f23d3eb")
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, 1, 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
import torch.nn.functional as F

torch.manual_seed(0)

# Create a small 4D tensor for image-like data (batch=2, channels=4, height=3, width=5)
x = torch.randn(2, 4, 3, 5)

# Create a scale factor (must be > 1 for downscaling)
scale = 2

ARGS = {
    'x': x,
    'scale': scale,
}
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, 1, 2) — shape (2, 4, 1, 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
b_spline_basis Ran
ran on a synthesized input
isosnovik/disco/models/basis/bspline_basis.py
code served (permissive licence) · get_code("7bd588bfb7a00300")
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 (9, 8, 8) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    "size": 8,
    "scale": 2.0,
    "effective_size": 3,
    "mult": 1.5,
    "order": 2
}
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 Tensor, ndim 3, dtype family float, shape (9, 8, 8) — shape (9, 8, 8) 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
cart2pol Ran
ran on a synthesized input
isosnovik/disco/models/basis/fb_basis.py
code served (permissive licence) · get_code("842820cbdd13d1ee")
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, ndim 2, 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

# Create small numpy arrays for x and y parameters
x = np.array([3.0, 4.0])
y = np.array([4.0, 3.0])

ARGS = {
    'x': x,
    'y': y
}
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, ndim 2, dtype family float — a tuple 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
dilated_standard_basis Ran
ran on a synthesized input
isosnovik/disco/models/basis/disco.py
code served (permissive licence) · get_code("f2822a43db40ecdc")
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 (49, 15, 15) 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)

# Analyze the function:
# - size: int, used in padding calculation
# - effective_size: int, used in formula for _size
# - scale: int, used for scaling and indexing
# All parameters are integers used in mathematical operations

ARGS = {
    "size": 16,
    "effective_size": 7,
    "scale": 2
}
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 (49, 15, 15) — shape (49, 15, 15) 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
fb_basis_grid Ran
ran on a synthesized input
isosnovik/disco/models/basis/fb_basis.py
code served (permissive licence) · get_code("2f6a43de5baa4afb")
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 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 os
import numpy as np
from scipy import special
import scipy.io
import tempfile

# Create a temporary bessel.mat file for the function to load
temp_dir = tempfile.mkdtemp()
bessel_path = os.path.join(temp_dir, 'bessel.mat')

# Create mock bessel data with the expected structure
# bessel array has columns: [ang_freq, rad_freq, eigenvalue, max_radius]
bessel_data = np.array([
    [0, 1, 2.4048, 2.4048],
    [0, 2, 5.5201, 5.5201],
    [1, 1, 3.8317, 3.8317],
    [1, 2, 7.0156, 7.0156],
    [2, 1, 5.1356, 5.1356],
    [2, 2, 8.4172, 8.4172],
    [3, 1, 6.3802, 6.3802],
    [3, 2, 9.7610, 9.7610],
], dtype=np.float64)

scipy.io.savemat(bessel_path, {'bessel': bessel_data})

# Monkey-patch the BESSEL_PATH in the function's module
import sys
import types

# Create a mock module with the bessel.mat path set to our temp file
mock_module = types.ModuleType('mock_fb')
mock_module.BESSEL_PATH = bessel_path

ARGS = {
    'size': 16,
    'scale': 1.0,
    'num_funcs': 8,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, 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 float — no output shape is stated: observe_signature.py could not observe it (import:NameError)
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_basis_filename Ran
ran on a synthesized input
isosnovik/disco/models/basis/disco.py
code served (permissive licence) · get_code("5c978437cf8b0c96")
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 str. 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

torch.manual_seed(0)

ARGS = {
    'size': 256,
    'effective_size': 128,
    'scales': [0.5, 1.0, 2.0, 4.5]
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_num_parameters Ran
ran on a synthesized input
isosnovik/disco/utils/model_utils.py
code served (permissive licence) · get_code("623393a11de5dcee")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create a simple nn.Module with some trainable parameters
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

module = SimpleModel()

ARGS = {
    "module": module
}
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 int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalize_basis_by_min_scale Ran
ran on a synthesized input
isosnovik/disco/models/basis/base.py
code served (permissive licence) · get_code("76bab8a4f1d47111")
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, 5) 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 as nn

torch.manual_seed(0)

# The function expects 'basis' which is used as:
# - basis.pow(2) - needs to support pow method
# - .sum([2, 3], keepdim=True) - needs dimensions 2 and 3
# - .sqrt() - needs to support sqrt
# - [:, [0]] - needs indexing with first dimension and column 0
# - / norm - needs to support division
# This is a tensor operation, so basis should be a torch.Tensor

# From the indexing [:, [0]], basis needs at least 2 dimensions
# From .sum([2, 3], ...), basis needs at least 4 dimensions
# Shape should be (batch, channels, height, width) typically
# Let's use batch=2, channels=4, height=3, width=5 for distinct sizes

basis = torch.randn(2, 4, 3, 5)

ARGS = {
    "basis": basis
}
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, 5) — shape (2, 4, 3, 5) 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
ses_max_projection Ran
ran on a synthesized input
isosnovik/disco/models/ses_conv.py
code served (permissive licence) · get_code("6aebfa2762d63d7a")
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)

# The function takes a tensor x and calls x.max(2)[0]
# This means x should be a tensor with at least 3 dimensions
# max(2) gets the max along dimension 2, returning (values, indices)
# [0] extracts just the values
# Result shape will be x.shape with dimension 2 removed

x = torch.randn(2, 4, 3)  # batch=2, dim1=4, dim2=3

ARGS = {
    "x": x
}
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
steerable Ran
ran on a synthesized input
isosnovik/disco/models/basis/fb_basis.py
code served (permissive licence) · get_code("a247e3f722fe0459")
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; 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 os
import numpy as np
from scipy import special
import scipy.io
import torch
import torch.nn.functional as F
from unittest.mock import MagicMock, patch

torch.manual_seed(0)
np.random.seed(0)

# Create a mock bessel.mat file in memory
# The function expects bessel to be a numpy array with shape (N, 4)
# Columns: [angular_freq, radial_freq, R_n, freq_limit]
mock_bessel_data = np.array([
    [0, 0, 0.5, 0.3],
    [1, 1, 1.2, 0.5],
    [1, 2, 1.8, 0.7],
    [2, 1, 2.1, 0.9],
    [2, 2, 2.5, 1.1],
    [3, 1, 3.0, 1.3],
    [3, 2, 3.4, 1.5],
    [4, 1, 4.0, 1.7],
], dtype=np.float64)

# Patch scipy.io.loadmat to return our mock data
original_loadmat = scipy.io.loadmat

def mock_loadmat(path):
    if 'bessel.mat' in path:
        return {'bessel': mock_bessel_data}
    return original_loadmat(path)

# Apply the patch globally
scipy.io.loadmat = mock_loadmat

ARGS = {
    'size': 16,
    'scales': [1.0, 2.0],
    'effective_size': 3,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, 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 — no output shape is stated: observe_signature.py could not observe it (import:NameError)
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
steerable_cropped Ran
ran on a synthesized input
isosnovik/disco/models/basis/bspline_basis.py
code served (permissive licence) · get_code("bb40d6eb58dae47b")
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 (16, 3, 31, 31) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    'size': 32,
    'scales': [0.5, 1.0, 2.0],
    'effective_size': 4,
    'mult': 1.0,
    'order': 2
}
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 Tensor, ndim 4, dtype family float, shape (16, 3, 31, 31) — shape (16, 3, 31, 31) 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
b_spline_at Not yet run
unverified (harvested, no recorded run)
isosnovik/disco/models/basis/bspline_basis.py
code served (permissive licence) · get_code("046245e838fc3755")
How it ran

no invocation recorded.

mnist_bspline Not yet run
unverified (harvested, no recorded run)
isosnovik/disco/models/mnist_models.py
code served (permissive licence) · get_code("05b3e992b436ef66")
How it ran

no invocation recorded.

mnist_disco Not yet run
unverified (harvested, no recorded run)
isosnovik/disco/models/mnist_disco_models.py
code served (permissive licence) · get_code("7d110a06509b4080")
How it ran

no invocation recorded.

mnist_hermite Not yet run
unverified (harvested, no recorded run)
isosnovik/disco/models/mnist_models.py
code served (permissive licence) · get_code("63f99bb41c85c9b6")
How it ran

no invocation recorded.

mnist_lr_harmonics Not yet run
unverified (harvested, no recorded run)
isosnovik/disco/models/mnist_models.py
code served (permissive licence) · get_code("4cf2c2e420d24717")
How it ran

no invocation recorded.

wrn_disco Not yet run
unverified (harvested, no recorded run)
isosnovik/disco/models/stl_disco_models.py
code served (permissive licence) · get_code("6e52493a7bbe4933")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 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 17 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

Scale is often seen as a given, disturbing factor in many vision tasks. When doing so it is one of the factors why we need more data during learning. In recent work scale equivariance was added to convolutional neural networks. It was shown to be effective for a range of tasks. We aim for accurate scale-equivariant convolutional neural networks (SE-CNNs) applicable for problems where high granularity of scale and small kernel sizes are required. Current SE-CNNs rely on weight sharing and kernel rescaling, the latter of which is accurate for integer scales only. To reach accurate scale equivariance, we derive general constraints under which scale-convolution remains equivariant to discrete rescaling. We find the exact solution for all cases where it exists, and compute the approximation for the rest. The discrete scale-convolution pays off, as demonstrated in a new state-of-the-art classification on MNIST-scale and on STL-10 in the supervised learning setting. With the same SE scheme, we also improve the computational effort of a scale-equivariant Siamese tracker on OTB-13.

For agents

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

get_harvested_code_for_paper("2106.02733")
get_code_for_paper("2106.02733")
have("2106.02733")

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

curl https://syntology.ai/api/ran/2106.02733.json

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

[![Syntology run record](https://syntology.ai/api/ran/2106.02733.svg)](https://syntology.ai/paper/2106.02733)

Connect an agent — have() is free.