SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1608.00859 · 2016

Temporal Segment Networks: Towards Good Practices for Deep Action Recognition

arXiv · PDF · Open in the Atlas

Code that ran

10 of 24 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; 2 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 0 failures of our instrument; 14 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/1608.00859.json · badge: /api/ran/1608.00859.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
CrazySherman/goodlife community (classifier label)
classifier basis not recorded on the link
5 ran · 3 no recorded run
ZJCV/Non-local community (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
nhannguyen2709/video_recognition reimplementation (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
ZJCV/TSN community (classifier label)
classifier basis not recorded on the link
1 ran · 0 no recorded run
MichiganCOG/M-PACT community (classifier label)
classifier basis not recorded on the link
0 ran · 9 no recorded run
WavesUR/embedded_TSM reimplementation (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
rijuldhir/TSM reimplementation (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
convert_sync_bn Ran
ran on a synthesized input
ZJCV/Non-local/tsn/model/norm_helper.py
code served (permissive licence) · get_code("88828f6a43c2063e")
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 SimpleModel. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Build a simple model with BatchNorm layers that can be converted to SyncBatchNorm
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(3, 16, kernel_size=3)
        self.bn1 = nn.BatchNorm2d(16)
        self.conv2 = nn.Conv2d(16, 32, kernel_size=3)
        self.bn2 = nn.BatchNorm2d(32)
        self.fc = nn.Linear(32, 10)
    
    def forward(self, x):
        x = self.bn1(self.conv1(x))
        x = self.bn2(self.conv2(x))
        return x

model = SimpleModel()

# Create a process group (None is acceptable for single-process CPU usage)
process_group = None

ARGS = {
    "model": model,
    "process_group": process_group
}
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 SimpleModel — a SimpleModel output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleModel 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_norm Ran
ran on a synthesized input
ZJCV/Non-local/tsn/model/norm_helper.py
code served (permissive licence) · get_code("9092b8b72f48621b")
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 type. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from functools import partial
from types import SimpleNamespace

torch.manual_seed(0)

# Define a minimal GroupNormWrapper class since it's referenced in the function
class GroupNormWrapper(nn.Module):
    def __init__(self, num_channels, num_groups):
        super().__init__()
        self.norm = nn.GroupNorm(num_groups, num_channels)
    
    def forward(self, x):
        return self.norm(x)

# Create a minimal config object with the required structure
cfg = SimpleNamespace()
cfg.MODEL = SimpleNamespace()
cfg.MODEL.NORM = SimpleNamespace()

# Set up for BatchNorm2d test case
cfg.MODEL.NORM.TYPE = 'BatchNorm2d'
cfg.MODEL.NORM.GROUPS = 8

ARGS = {
    'cfg': cfg
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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_norm Ran
ran on a synthesized input
ZJCV/TSN/tsn/model/norm_helper.py
code served (permissive licence) · get_code("91b0ded288fc9495")
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 type. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from functools import partial

torch.manual_seed(0)

class GroupNormWrapper(nn.Module):
    def __init__(self, num_channels, num_groups):
        super().__init__()
        self.norm = nn.GroupNorm(num_groups, num_channels)
    
    def forward(self, x):
        return self.norm(x)

class MockCfgNode:
    def __init__(self, norm_type, num_groups=None):
        self.MODEL = MockModel(norm_type, num_groups)

class MockModel:
    def __init__(self, norm_type, num_groups=None):
        self.NORM = MockNorm(norm_type, num_groups)

class MockNorm:
    def __init__(self, norm_type, num_groups=None):
        self.TYPE = norm_type
        self.GROUPS = num_groups if num_groups is not None else 8

cfg_batchnorm = MockCfgNode('BatchNorm2d')
cfg_groupnorm = MockCfgNode('GroupNorm', num_groups=8)

ARGS = {
    'cfg': cfg_batchnorm
}
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 type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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
rgb_oversample Ran
ran on a synthesized input
CrazySherman/goodlife/pyActionRecog/utils/io.py
code served (permissive licence) · get_code("dc1ea541a1f247e4")
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 4, dtype family float, shape (10, 10, 8, 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

# Create a small RGB image (H=20, W=16, K=3 channels)
# Using distinct dimensions for clarity
image = np.random.rand(20, 16, 3).astype(np.float32)

# Define crop dimensions (height=10, width=8)
crop_dims = (10, 8)

ARGS = {
    "image": image,
    "crop_dims": crop_dims,
}
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 4, dtype family float, shape (10, 10, 8, 3) — shape (10, 10, 8, 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
rgb_to_parrots Ran
ran on a synthesized input
CrazySherman/goodlife/pyActionRecog/utils/io.py
code served (permissive licence) · get_code("0787dc11cf03c6bd")
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 4, dtype family float, shape (10, 3, 16, 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 numpy as np

# Create a small RGB image (H x W x K format)
# Using distinct dimensions: 32x24x3 (height=32, width=24, channels=3)
image = np.random.randint(0, 256, size=(32, 24, 3), dtype=np.uint8).astype(np.float32)

# Define crop dimensions (height, width)
crop_size = (16, 12)

ARGS = {
    'frame': image,
    'oversample': True,
    'mean_val': [104, 117, 123],
    'crop_size': crop_size,
}
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 4, dtype family float, shape (10, 3, 16, 12) — shape (10, 3, 16, 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
schedule Ran
ran, honoured the contract we drafted
nhannguyen2709/video_recognition/train_tsn_spatial_stream.py
pointer only (licence: NONE) · get_code("fa89677d377778f8")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, nhannguyen2709/video_recognition 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.

schedule Ran
ran, honoured the contract we drafted
nhannguyen2709/video_recognition/train_tsn_motion_stream.py
pointer only (licence: NONE) · get_code("5e212072c1ba4c65")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, nhannguyen2709/video_recognition 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.

softmax Ran
ran on a synthesized input
CrazySherman/goodlife/pyActionRecog/utils/metrics.py
code served (permissive licence) · get_code("a59fc47a6706506b")
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 (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 numpy as np

# Analyze the function:
# - raw_score: numpy array that supports .max(axis=-1) and arithmetic operations
# - T: scalar (default=1), used in multiplication
# The function performs softmax with temperature scaling on the last axis

# Create a small test case
raw_score = np.random.randn(2, 4)  # batch_size=2, num_classes=4

T = 1.0

ARGS = {
    "raw_score": raw_score,
    "T": T,
}
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 (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
top_k_acc Ran
ran on a synthesized input
CrazySherman/goodlife/pyActionRecog/utils/metrics.py
code served (permissive licence) · get_code("71665e40051d22da")
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 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 numpy as np

# Analyze the function:
# - lb_set: needs .intersection() method, so it's a set
# - scores: used with np.argsort(), so it's array-like
# - k: integer parameter with default value 3

# Create a small set of label indices
lb_set = {0, 2, 4}

# Create a scores array with some values
# Using more elements than k so argsort makes sense
scores = np.array([0.1, 0.5, 0.3, 0.9, 0.2, 0.7])

# k parameter (using default value)
k = 3

ARGS = {
    "lb_set": lb_set,
    "scores": scores,
    "k": k
}
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 int — 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
top_k_hit Ran
ran on a synthesized input
CrazySherman/goodlife/pyActionRecog/utils/metrics.py
code served (permissive licence) · get_code("5b4ba07c7dc78f0e")
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 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 numpy as np

# Analyze the function:
# - lb_set: needs .intersection() method, so it's a set-like object
# - scores: used with np.argsort(), so it's array-like
# - k: integer parameter

# Create a small set of indices to use as lb_set
lb_set = {0, 2, 4}

# Create scores array with some values
# The function will argsort and take top k indices
# We'll make it small but with enough elements
scores = np.array([10.5, 2.3, 7.1, 15.2, 3.8, 9.1])

k = 3

ARGS = {
    "lb_set": lb_set,
    "scores": scores,
    "k": k
}
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 int — 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
flow_stack_oversample Not yet run
unverified (harvested, no recorded run)
CrazySherman/goodlife/pyActionRecog/utils/io.py
code served (permissive licence) · get_code("7feeb18d1dbe89d8")
How it ran

no invocation recorded.

make_temporal_shift Not yet run
unverified (harvested, no recorded run)
WavesUR/embedded_TSM/ops/temporal_shift.py
pointer only (licence: NONE) · get_code("690432aed6251af7")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, WavesUR/embedded_TSM 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.

make_temporal_shift Not yet run
unverified (harvested, no recorded run)
rijuldhir/TSM/ops/temporal_shift.py
code served (permissive licence) · get_code("ea903b20756dc01d")
How it ran

no invocation recorded.

parse_armlc_splits Not yet run
unverified (harvested, no recorded run)
CrazySherman/goodlife/pyActionRecog/armlc.py
code served (permissive licence) · get_code("324b251cf5317ec4")
How it ran

no invocation recorded.

parse_label Not yet run
unverified (harvested, no recorded run)
CrazySherman/goodlife/pyActionRecog/armlc.py
code served (permissive licence) · get_code("7225c9c33010b8d9")
How it ran

no invocation recorded.

preprocess_for_eval Not yet run
unverified (harvested, no recorded run)
MichiganCOG/M-PACT/models/c3d/default_preprocessing.py
code served (permissive licence) · get_code("541b84f358db0382")
How it ran

no invocation recorded.

preprocess_for_eval Not yet run
unverified (harvested, no recorded run)
MichiganCOG/M-PACT/models/c3d/tf_version_HMDB51_preprocessing.py
code served (permissive licence) · get_code("2d6a9afae20381b6")
How it ran

no invocation recorded.

preprocess_for_eval Not yet run
unverified (harvested, no recorded run)
MichiganCOG/M-PACT/models/i3d/default_preprocessing.py
code served (permissive licence) · get_code("634caffb34a18a9c")
How it ran

no invocation recorded.

preprocess_for_train Not yet run
unverified (harvested, no recorded run)
MichiganCOG/M-PACT/models/c3d/default_preprocessing.py
code served (permissive licence) · get_code("1076f344b59ea48f")
How it ran

no invocation recorded.

preprocess_for_train Not yet run
unverified (harvested, no recorded run)
MichiganCOG/M-PACT/models/c3d/tf_version_HMDB51_preprocessing.py
code served (permissive licence) · get_code("5209697119958449")
How it ran

no invocation recorded.

preprocess_for_train Not yet run
unverified (harvested, no recorded run)
MichiganCOG/M-PACT/models/i3d/default_preprocessing.py
code served (permissive licence) · get_code("684ff8396e14fded")
How it ran

no invocation recorded.

preprocess_image Not yet run
unverified (harvested, no recorded run)
MichiganCOG/M-PACT/models/c3d/default_preprocessing.py
code served (permissive licence) · get_code("f015c09b9fe3403d")
How it ran

no invocation recorded.

preprocess_image Not yet run
unverified (harvested, no recorded run)
MichiganCOG/M-PACT/models/c3d/tf_version_HMDB51_preprocessing.py
code served (permissive licence) · get_code("86d341d4d8e50c6e")
How it ran

no invocation recorded.

preprocess_image Not yet run
unverified (harvested, no recorded run)
MichiganCOG/M-PACT/models/i3d/default_preprocessing.py
code served (permissive licence) · get_code("b9dd5244dad95ed4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 24 samples lifted from this paper's linked repositories, 24 carry a function identity and 0 do not. Of the 23 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 22 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
make_temporal_shift 2 other papers
2 later
1811.08383 (2018-11)
TSM: Temporal Shift Module for Efficient Video Understanding
2203.13611 (2022-03)
Class-Incremental Learning for Action Recognition in Videos

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

Deep convolutional networks have achieved great success for visual recognition in still images. However, for action recognition in videos, the advantage over traditional methods is not so evident. This paper aims to discover the principles to design effective ConvNet architectures for action recognition in videos and learn these models given limited training samples. Our first contribution is temporal segment network (TSN), a novel framework for video-based action recognition. which is based on the idea of long-range temporal structure modeling. It combines a sparse temporal sampling strategy and video-level supervision to enable efficient and effective learning using the whole action video. The other contribution is our study on a series of good practices in learning ConvNets on video data with the help of temporal segment network. Our approach obtains the state-the-of-art performance on the datasets of HMDB51 ( $ 69.4\% $) and UCF101 ($ 94.2\% $). We also visualize the learned ConvNet models, which qualitatively demonstrates the effectiveness of temporal segment network and the proposed good practices.

For agents

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

get_harvested_code_for_paper("1608.00859")
get_code_for_paper("1608.00859")
have("1608.00859")

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

curl https://syntology.ai/api/ran/1608.00859.json

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

[![Syntology run record](https://syntology.ai/api/ran/1608.00859.svg)](https://syntology.ai/paper/1608.00859)

Connect an agent — have() is free.