SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2110.00529 · NeurIPS · 2021

Unsupervised Motion Representation Learning with Capsule Autoencoders

Mohan Kankanhalli, Yongkang Wong, Ziwei Xu, Xudong Shen

arXiv · PDF · Open in the Atlas

Code that ran

15 of 20 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (4 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 2 violated it and 5 have no contract verdict; 7 failures of our instrument; 5 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/2110.00529.json · badge: /api/ran/2110.00529.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
ziweixu/capsulemotion canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
ZiweiXU/CapsuleMotion marked official in Papers with Code
Papers with Code archive
14 ran · 5 no recorded run
FunctionStatusWhere it lives
BatchLinear Ran
ran on a synthesized input · constructed an object; no result was computed
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("791124774ac72001")
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 an instance of its own class, BatchLinear: the run constructed the object and did not call it, so no result was computed. 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
from torch import nn
import math

torch.manual_seed(0)

ARGS = {
    "in_dim": 4,
    "out_dim": 3,
    "indep_dims": [2],
    "use_bias": 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 BatchLinear — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a BatchLinear 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
SegmentDecoder Ran
ran on a synthesized input · constructed an object; no result was computed
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("67a918cc07243e1d")
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 an instance of its own class, SegmentDecoder: the run constructed the object and did not call it, so no result was computed. 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
from torch import nn
import numpy as np

torch.manual_seed(0)

ARGS = {
    "num_sni_temp": 16,
    "len_sni_temp": 8,
    "num_seg_temp": 48,
    "num_seg_param": 7,
    "max_len": 64,
    "warping_dim": 3,
    "constrain_scale": 1.0,
    "constrain_trans": 1.5,
}
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 SegmentDecoder — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a SegmentDecoder 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
SegmentEncoder Ran
ran on a synthesized input · constructed an object; no result was computed
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("4e176ccdd110ccc3")
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 an instance of its own class, SegmentEncoder: the run constructed the object and did not call it, so no result was computed. 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
from torch import nn
import math

torch.manual_seed(0)

ARGS = {
    "num_sni_temp": 4,
    "len_sni_temp": 2,
    "num_seg_temp": 3,
    "max_len": 8,
    "warping_dim": 3,
    "num_seg_param": 7,
    "lstm_hidden": 16,
    "fc_hidden": 8,
}
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 SegmentEncoder — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a SegmentEncoder 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
SnippetEncoder Ran
ran on a synthesized input · constructed an object; no result was computed
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("9d5043f4be759579")
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 an instance of its own class, SnippetEncoder: the run constructed the object and did not call it, so no result was computed. 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
from torch import nn
import numpy as np

torch.manual_seed(0)

ARGS = {
    "num_sni_temp": 16,
    "len_sni_temp": 8,
    "intm_channels": 8,
    "max_len": 64,
    "num_sni_param": 5,
    "num_object": 1,
    "coor_dim": 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 SnippetEncoder — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a SnippetEncoder 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
compact_pose Ran
ran, violated the contract we drafted
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("6d4d34062e1c5530")
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, 12) 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
from torch.autograd import Variable

torch.manual_seed(0)

# compact_pose expects a full_pose tensor of shape (B, 3, 3) or (B, 4, 4)
# Looking at the function body:
# - It checks if full_pose.shape[-1] == 4 (for 3D case)
# - If true, calls compact_pose_3d which expects (B, 4, 4)
# - If false, processes (B, 3, 3) case
# Let's test with a (B, 4, 4) tensor to exercise the 3D path

batch_size = 2
full_pose_tensor = torch.randn(batch_size, 4, 4)

ARGS = {
    "full_pose": full_pose_tensor
}
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, 12) — shape (2, 12) 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
compact_pose_3d Ran
ran, violated the contract we drafted
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("2e5f0a62fa71eea9")
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, 12) 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
from torch.autograd import Variable

torch.manual_seed(0)

# Create a small (B, 4, 4) transformation matrix for testing compact_pose_3d
# Batch size 2, so we have 2 transformation matrices
batch_size = 2
full_pose_tensor = torch.randn(batch_size, 4, 4)

ARGS = {
    "full_pose": full_pose_tensor
}
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, 12) — shape (2, 12) 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
constrain_pose Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("626bb1811a1b7445")
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, 6) 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)

# Test constrain_pose with different pose dimensions
# The function supports pose shapes of (..., 3), (..., 6), or (..., 12)

# Create a small batch of pose vectors with dimension 6
# Shape: (2, 6) - batch size 2, pose dimension 6
pose_6d = torch.randn(2, 6)

ARGS = {
    "pose": pose_6d,
    "scale": 1.0,
    "trans_range": 1.5,
}
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 Tensor, ndim 2, dtype family float, shape (2, 6) — shape (2, 6) 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
constrain_pose_3d Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("e1661d691e4a901a")
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, 12) 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 a pose tensor of shape (2, 12)
# The function expects (..., 12) where the last dimension contains:
# [a11, a12, a13, tx, a21, a22, a23, ty, a31, a32, a33, tz]
pose = torch.randn(2, 12)

# Define arguments for the function
ARGS = {
    "pose": pose,
    "scale": 1.0,
    "trans_range": 1.5,
}
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 Tensor, ndim 2, dtype family float, shape (2, 12) — shape (2, 12) 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
expand_pose Ran
ran; our contract draft was wrong, not the code
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("9efa4b606d03159d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (2, 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
from torch.autograd import Variable

torch.manual_seed(0)

# Create a small pose tensor of shape N x 3
# N=2 (batch size), 3 (s, x, y dimensions)
pose = torch.randn(2, 3)

ARGS = {
    "pose": pose
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 2, 3) — shape (2, 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
fixed_template_initializer Ran
ran, honoured the contract we drafted
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("bf75b48517195b2b")
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 (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)

ARGS = {
    'num_sni_temp': 4,
    'len_sni_temp': 3,
    'scale_factor': 1.0,
    'mode': '2d'
}
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 (4, 3, 3) — shape (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
full_pose Ran
ran; our contract draft was wrong, not the code
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("6b422476f7a75d45")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (2, 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
from torch.autograd import Variable

torch.manual_seed(0)

# Test case 1: pose with shape (2, 3)
pose_3d = torch.randn(2, 3)

ARGS = {
    "pose": pose_3d
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 3, 3) — shape (2, 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
full_pose_3d Ran
ran; our contract draft was wrong, not the code
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("dbac1fbe39d62402")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (2, 4, 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 full_pose_3d takes a pose tensor of shape (B, 12)
# and expands it to (B, 4, 4) transformation matrix
# B is batch size, 12 flattens to 3x4 matrix before adding last row

# Create a small batch of pose data
# Shape: (batch_size=2, pose_dim=12)
pose = torch.randn(2, 12)

ARGS = {
    "pose": pose
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 4, 4) — shape (2, 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
init_truncated_normal Ran
ran; our contract draft was wrong, not the code
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("164fc00eb042e928")
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 Parameter, 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 torch
import math
from torch.nn.init import _calculate_fan_in_and_fan_out

torch.manual_seed(0)

# Create a parameter tensor with a realistic shape for initialization
# Using shape (4, 3) to have distinct dimensions
param = torch.nn.Parameter(torch.empty(4, 3))

ARGS = {
    "param": param,
    "stdv": None,
    "fan_in": None,
}
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 Parameter, ndim 2, dtype family float — no output shape is stated: observe_signature.py recorded Tensor ndim 2, 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
loss_fn Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ziweixu/capsulemotion/lib/mcae/model.py
code served (permissive licence) · get_code("3aa0d94122193b9e")
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) and returned tuple, 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 torch
import torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

# Construct minimal tensors matching the function's expectations
batch_size = 2
seq_len = 8
feature_dim = 4
num_classes = 3
lSnT = 2  # len_sni_temp

# x: input tensor (batch, seq_len, feature_dim)
x = torch.randn(batch_size, seq_len, feature_dim)

# x_sni: snippet reconstruction (batch, seq_len, feature_dim)
x_sni = torch.randn(batch_size, seq_len, feature_dim)

# sni_pres: snippet presence (batch, some_dim)
sni_pres = torch.randn(batch_size, 5)

# seg_pres: segment presence (batch, some_dim) - used in contrastive loss
seg_pres = torch.randn(batch_size, 6)

# x_seg: segment reconstruction (batch, seq_len, feature_dim)
x_seg = torch.randn(batch_size, seq_len, feature_dim)

# sni_params: snippet parameters (batch, num_params, param_dim1, param_dim2)
sni_params = torch.randn(batch_size, 3, 2, 2)

# seg_params: segment parameters (batch, num_params, param_dim1, param_dim2)
seg_params = torch.randn(batch_size, 3, 2, 2)

# sni_params_trans: transformed snippet parameters
sni_params_trans = torch.randn(batch_size, 3, 2, 2)

# sni_params_recons: reconstructed snippet parameters
sni_params_recons = torch.randn(batch_size, 3, 2, 2)

# seg_lgts: segment logits (batch, num_classes)
seg_lgts = torch.randn(batch_size, num_classes)

# y: class labels (batch,)
y = torch.randint(0, num_classes, (batch_size,))

# results dict containing all required keys
results = {
    'x_sni': x_sni,
    'sni_pres': sni_pres,
    'seg_pres': seg_pres,
    'x_seg': x_seg,
    'sni_params': sni_params,
    'seg_params': seg_params,
    'sni_params_trans': sni_params_trans,
    'sni_params_recons': sni_params_recons,
    'seg_lgts': seg_lgts,
    'len_sni_temp': lSnT,
}

# loss_weights dict
loss_weights = {
    'sni': 1.0,
    'seg': 1.0,
    'cont': 1.0,
    'reg': 1.0,
    'con': 1.0,
    'cls': 1.0,
}

ARGS = {
    'x': x,
    'results': results,
    'is_valtest': False,
    'y': y,
    'loss_weights': loss_weights,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, 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
myself Ran
ran on a synthesized input
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("a5e44072bb9d9d5d")
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 inspect

def myself():
    return inspect.stack()[1][3]

# The function 'myself' uses inspect.stack()[1][3] to return the name of the calling function.
# It takes no parameters, so ARGS should be an empty dict.

ARGS = {}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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
MCAE Not yet run
unverified (harvested, no recorded run)
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("efbad1cbdacbe091")
How it ran

no invocation recorded.

Module Not yet run
unverified (harvested, no recorded run)
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("335cc3b992ddaf73")
How it ran

no invocation recorded.

SnippetDecoder Not yet run
unverified (harvested, no recorded run)
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("07fdb5a4536b7a31")
How it ran

no invocation recorded.

_truncated_normal Not yet run
unverified (harvested, no recorded run)
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("f7fcffe777530157")
How it ran

no invocation recorded.

fixed_template_initializer_3d Not yet run
unverified (harvested, no recorded run)
ZiweiXU/CapsuleMotion/lib/mcae/model.py
code served (permissive licence) · get_code("a393a223aba0b7cb")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 20 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 20 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 the Motion Capsule Autoencoder (MCAE), which addresses a key challenge in the unsupervised learning of motion representations: transformation invariance. MCAE models motion in a two-level hierarchy. In the lower level, a spatio-temporal motion signal is divided into short, local, and semanticagnostic snippets. In the higher level, the snippets are aggregated to form fulllength semantic-aware segments. For both levels, we represent motion with a set of learned transformation invariant templates and the corresponding geometric transformations by using capsule autoencoders of a novel design. This leads to a robust and efficient encoding of viewpoint changes. MCAE is evaluated on a novel Trajectory20 motion dataset and various real-world skeleton-based human action datasets. Notably, it achieves better results than baselines on Trajectory20 with considerably fewer parameters and state-of-the-art performance on the unsupervised skeleton-based action recognition task.

For agents

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

get_harvested_code_for_paper("2110.00529")
get_code_for_paper("2110.00529")
have("2110.00529")

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

curl https://syntology.ai/api/ran/2110.00529.json

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

[![Syntology run record](https://syntology.ai/api/ran/2110.00529.svg)](https://syntology.ai/paper/2110.00529)

Connect an agent — have() is free.