SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2201.02849 · 2022

Spatio-Temporal Tuples Transformer for Skeleton-Based Action Recognition

arXiv · PDF · Open in the Atlas

Code that ran

7 of 11 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 2 have no contract verdict; 5 failures of our instrument; 4 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/2201.02849.json · badge: /api/ran/2201.02849.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
heleiqiu/sttformer canonical (classifier label)
classifier basis not recorded on the link
7 ran · 4 no recorded run
FunctionStatusWhere it lives
denoising_by_length Ran
ran; our contract draft was wrong, not the code
heleiqiu/sttformer/gendata/ntu/get_raw_denoised_data.py
code served (permissive licence) · get_code("d63c7bd0fe3e1881")
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 logging

# Set up logging to avoid warnings
logging.basicConfig(level=logging.INFO)

# Define ARGS with valid arguments for denoising_by_length
ARGS = {
    'ske_name': 'skeleton_001',
    'bodies_data': {
        'body_1': {
            'interval': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],  # length 13, > threshold
            'motion': 0.75
        },
        'body_2': {
            'interval': [0, 1, 2],  # length 3, <= threshold (will be filtered)
            'motion': 0.25
        },
        'body_3': {
            'interval': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11],  # length 12, > threshold
            'motion': 0.50
        },
        'body_4': {
            'interval': [0],  # length 1, <= threshold (will be filtered)
            'motion': 0.10
        }
    }
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 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
denoising_by_spread Ran
ran; our contract draft was wrong, not the code
heleiqiu/sttformer/gendata/ntu/get_raw_denoised_data.py
code served (permissive licence) · get_code("d3358705b52bccbe")
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 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import logging

# Set up logging to avoid warnings
logging.basicConfig(level=logging.INFO)

# Create test data matching the function's expectations
# bodies_data is a dict with bodyID keys, each containing 'joints', 'interval', 'motion'

# Create joints data: shape should be (num_frames * 25, 3) before reshape to (-1, 25, 3)
# Let's create 4 frames of skeleton data (25 joints per frame, 3 coordinates each)
num_frames = 4
num_joints = 25
coords = 3

# Body 1: Valid data with good spread
joints_body1 = np.random.randn(num_frames * num_joints, coords).astype(np.float32)
# Ensure some spread variation in X and Y to make it valid
joints_body1[:, 0] = joints_body1[:, 0] * 0.5  # reduce X spread
joints_body1[:, 1] = joints_body1[:, 1] * 2.0  # increase Y spread

# Body 2: Noisy data with poor spread (should be filtered out)
joints_body2 = np.random.randn(num_frames * num_joints, coords).astype(np.float32)
joints_body2[:, 0] = joints_body2[:, 0] * 1.5  # X spread too large
joints_body2[:, 1] = joints_body2[:, 1] * 0.3  # Y spread too small

bodies_data = {
    'body1': {
        'joints': joints_body1,
        'interval': [0, 1, 2, 3],  # 4 frames
        'motion': 0.5
    },
    'body2': {
        'joints': joints_body2,
        'interval': [0, 1, 2, 3],  # 4 frames
        'motion': 2.0
    }
}

ske_name = 'test_skeleton_001'

ARGS = {
    'ske_name': ske_name,
    'bodies_data': bodies_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, 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
downsample Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
heleiqiu/sttformer/feeders/tools.py
pointer only (licence: MIT); the stored copy get_code serves, from liujf69/MMCL-Action, is pointer only (licence: NONE) · get_code("8b14974e8aa7320f")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from liujf69/MMCL-Action, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_valid_frames_by_spread Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
heleiqiu/sttformer/gendata/ntu/get_raw_denoised_data.py
code served (permissive licence) · get_code("785086bb1a2ced64")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family int; 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 numpy as np

torch.manual_seed(0)

# The function expects:
# - points: a tensor with shape (num_frames, num_joints, 2)
#   It indexes points[i, :, 0] for x coordinates and points[i, :, 1] for y coordinates
#   It calls .max() and .min() on these slices
#   So we need at least 2 frames, multiple joints, and 2 coordinate dimensions

points = torch.randn(2, 4, 2)

ARGS = {
    "points": points
}
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 list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
remove_nan_frames Ran
ran on a synthesized input
heleiqiu/sttformer/gendata/ntu/seq_transformation.py
code served (permissive licence) · get_code("1cc3adc4a2068ead")
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, 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 numpy as np
import logging

# Create a simple logger that captures NaN information
nan_logger = logging.getLogger('nan_logger')
nan_logger.setLevel(logging.INFO)

# Create a skeleton name for the skeleton
ske_name = "test_skeleton_001"

# Create skeleton joints array with shape (num_frames, num_joints, num_coordinates)
# Mix of valid frames and frames with NaN values
ske_joints = np.array([
    [[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]],  # frame 0: valid
    [[np.nan, 2.0, 3.0], [4.0, 5.0, 6.0]],  # frame 1: has NaN
    [[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]],  # frame 2: valid
    [[1.0, np.nan, 3.0], [np.nan, 5.0, 6.0]],  # frame 3: has NaN
], dtype=np.float32)

ARGS = {
    'ske_name': ske_name,
    'ske_joints': ske_joints,
    'nan_logger': nan_logger,
}
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 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
seq_translation Ran
ran on a synthesized input
heleiqiu/sttformer/gendata/ntu/seq_transformation.py
code served (permissive licence) · get_code("477b711e99221772")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list. 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

# Create sample skeleton joint data
# The function expects a list/array of skeleton sequences
# Each sequence has shape (num_frames, num_joints)
# num_joints can be 75 (1 body) or 150 (2 bodies)

# Create a simple test case with 1 body (75 joints)
num_frames_1 = 10
num_joints_1 = 75
skes_joints_1body = np.random.randn(num_frames_1, num_joints_1).astype(np.float32)
# Ensure first frame has non-zero values in first 75 joints (for origin calculation)
skes_joints_1body[0, :75] = np.random.randn(75) + 5.0

# Create a test case with 2 bodies (150 joints)
num_frames_2 = 12
num_joints_2 = 150
skes_joints_2bodies = np.random.randn(num_frames_2, num_joints_2).astype(np.float32)
# Ensure first frame has non-zero values (for origin calculation)
skes_joints_2bodies[0, :75] = np.random.randn(75) + 5.0
skes_joints_2bodies[0, 75:] = np.random.randn(75) + 5.0
# Add some frames with missing data (all zeros) for body 1
skes_joints_2bodies[3, :75] = 0
skes_joints_2bodies[5, :75] = 0
# Add some frames with missing data (all zeros) for body 2
skes_joints_2bodies[4, 75:] = 0

# Create list of skeleton sequences
skes_joints = [skes_joints_1body, skes_joints_2bodies]

ARGS = {
    "skes_joints": skes_joints
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list 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
valid_crop_resize Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
heleiqiu/sttformer/feeders/tools.py
pointer only (licence: MIT); the stored copy get_code serves, from liujf69/MMCL-Action, is pointer only (licence: NONE) · get_code("37939bae213ecfd4")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from liujf69/MMCL-Action, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

frame_translation Not yet run
unverified (harvested, no recorded run)
heleiqiu/sttformer/gendata/ntu/seq_transformation.py
code served (permissive licence) · get_code("5ccb28010b540cbf")
How it ran

no invocation recorded.

get_raw_bodies_data Not yet run
unverified (harvested, no recorded run)
heleiqiu/sttformer/gendata/ntu/get_raw_skes_data.py
code served (permissive licence) · get_code("5b01cb771784850f")
How it ran

no invocation recorded.

get_raw_bodies_data Not yet run
unverified (harvested, no recorded run)
heleiqiu/sttformer/gendata/ntu120/get_raw_skes_data.py
code served (permissive licence) · get_code("a77a4bd849c26036")
How it ran

no invocation recorded.

temporal_slice Not yet run
unverified (harvested, no recorded run)
heleiqiu/sttformer/feeders/tools.py
pointer only (licence: MIT); the stored copy get_code serves, from liujf69/MMCL-Action, is pointer only (licence: NONE) · get_code("2f7f389f0584c97e")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from liujf69/MMCL-Action, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 6 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 0 the same month, 3 later) and 5 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
downsample 10 other papers
1 dated earlier, 8 later, 1 same year, order not determinable
1904.12659 (2019-04)
Actional-Structural Graph Convolutional Networks for Skeleton-based Action Recognition
ijcai2022_0511 (2022 (year only))
EGCN: An Ensemble-based Learning Framework for Exploring Effective Skeleton-based Rehabilitation Exercise Assessment
2208.10741 (2022-08)
Hierarchically Decomposed Graph Convolutional Networks for Skeleton-Based Action Recognition
and 7 more not named here
temporal_slice 10 other papers
1 dated earlier, 8 later, 1 same year, order not determinable
1904.12659 (2019-04)
Actional-Structural Graph Convolutional Networks for Skeleton-based Action Recognition
ijcai2022_0511 (2022 (year only))
EGCN: An Ensemble-based Learning Framework for Exploring Effective Skeleton-based Rehabilitation Exercise Assessment
2208.10741 (2022-08)
Hierarchically Decomposed Graph Convolutional Networks for Skeleton-Based Action Recognition
and 7 more not named here
valid_crop_resize 7 other papers
7 later
2208.10741 (2022-08)
Hierarchically Decomposed Graph Convolutional Networks for Skeleton-Based Action Recognition
2211.09590 (2022-11)
Hypergraph Transformer for Skeleton-based Action Recognition
2305.11468 (2023-05)
Overcoming Topology Agnosticism: Enhancing Skeleton-Based Action Recognition through Redefined Skeletal Topology Awareness
and 4 more not named here
denoising_by_length 1 other paper
1 dated earlier
1904.01189 (2019-04)
Semantics-Guided Neural Networks for Efficient Skeleton-Based Human Action Recognition
denoising_by_spread 1 other paper
1 dated earlier
1904.01189 (2019-04)
Semantics-Guided Neural Networks for Efficient Skeleton-Based Human Action Recognition
get_valid_frames_by_spread 1 other paper
1 dated earlier
1904.01189 (2019-04)
Semantics-Guided Neural Networks for Efficient Skeleton-Based Human Action Recognition

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

Capturing the dependencies between joints is critical in skeleton-based action recognition task. Transformer shows great potential to model the correlation of important joints. However, the existing Transformer-based methods cannot capture the correlation of different joints between frames, which the correlation is very useful since different body parts (such as the arms and legs in "long jump") between adjacent frames move together. Focus on this problem, A novel spatio-temporal tuples Transformer (STTFormer) method is proposed. The skeleton sequence is divided into several parts, and several consecutive frames contained in each part are encoded. And then a spatio-temporal tuples self-attention module is proposed to capture the relationship of different joints in consecutive frames. In addition, a feature aggregation module is introduced between non-adjacent frames to enhance the ability to distinguish similar actions. Compared with the state-of-the-art methods, our method achieves better performance on two large-scale datasets.

For agents

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

get_harvested_code_for_paper("2201.02849")
get_code_for_paper("2201.02849")
have("2201.02849")

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

curl https://syntology.ai/api/ran/2201.02849.json

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

[![Syntology run record](https://syntology.ai/api/ran/2201.02849.svg)](https://syntology.ai/paper/2201.02849)

Connect an agent — have() is free.