SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1906.04236 · 2019

Identifying Visible Actions in Lifestyle Vlogs

arXiv · PDF · Open in the Atlas

Code that ran

9 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 9 have no contract verdict; 0 failures of our instrument; 2 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/1906.04236.json · badge: /api/ran/1906.04236.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
MichiganNLP/vlog_action_recognition canonical (classifier label)
classifier basis: census
9 ran · 2 no recorded run
FunctionStatusWhere it lives
agreement_total Ran
ran on a synthesized input
MichiganNLP/vlog_action_recognition/amt/compute_agreement.py
code served (permissive licence) · get_code("add9845de22feb70")
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 float, ndim 0, dtype family float, 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 numpy as np

torch.manual_seed(0)

# The function takes dict_hit_agreement as input
# It accesses .values() and .keys() on the dict
# It expects numeric values that can be summed and averaged
# Based on the print statement, this appears to be agreement scores

dict_hit_agreement = {
    'hit_1': 0.95,
    'hit_2': 0.87,
    'hit_3': 0.92,
    'hit_4': 0.88,
}

ARGS = {
    'dict_hit_agreement': dict_hit_agreement
}
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 float, ndim 0, dtype family float, 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
build_dense_model Ran
ran on a synthesized input
MichiganNLP/vlog_action_recognition/classify/main_model.py
code served (permissive licence) · get_code("f634d81b68c16596")
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 Sequential. 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

# input_size is used as the input_dim parameter in the first Dense layer
# It should be a positive integer representing the number of input features
input_size = 10

ARGS = {
    'input_size': input_size
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
keras, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential 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
compute_agreement Ran
ran on a synthesized input
MichiganNLP/vlog_action_recognition/amt/read_csv_results.py
code served (permissive licence) · get_code("633a98e67cf8f4f7")
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 numpy as np
from sklearn.metrics import cohen_kappa_score, accuracy_score

# Test case 1: Two raters with binary ratings
rater1_two = [0, 1, 0, 1, 1, 0]
rater2_two = [0, 1, 1, 1, 0, 0]

# Test case 2: Three raters with binary ratings
rater1_three = [0, 1, 0, 1, 1, 0]
rater2_three = [0, 1, 1, 1, 0, 0]
rater3_three = [0, 1, 0, 1, 1, 0]

ARGS = {
    "rater1": rater1_two,
    "rater2": rater2_two,
    "rater3": None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — 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
compute_ratings Ran
ran on a synthesized input
MichiganNLP/vlog_action_recognition/amt/compute_agreement.py
code served (permissive licence) · get_code("ef2f2d82063b646a")
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 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 numpy as np
from collections import OrderedDict, Counter

# The function expects df_worker_1, df_worker_2, and optionally df_worker_3
# Looking at the function body, it:
# 1. Iterates through len(df_worker_1)
# 2. Accesses elements by index: df_worker_1[i], df_worker_2[i], df_worker_3[i]
# 3. Uses Counter on the list of results
# 4. Checks for counts of 0 and 1

# These appear to be list-like objects containing binary values (0 or 1)
# Let's create simple lists of binary values

df_worker_1 = [0, 1, 0, 1, 1]
df_worker_2 = [0, 0, 1, 1, 0]
df_worker_3 = [1, 1, 0, 1, 1]

ARGS = {
    'df_worker_1': df_worker_1,
    'df_worker_2': df_worker_2,
    'df_worker_3': df_worker_3
}
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 2, dtype family float — 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
create_model_concat Ran
ran on a synthesized input
MichiganNLP/vlog_action_recognition/classify/main_model.py
code served (permissive licence) · get_code("1a9424d013afe82b")
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 Functional. 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 keras
from keras.layers import Dense, Dropout, LSTM, Input
from keras import backend as K, Model

# input_dim_video is used as: Input(shape=(input_dim_video[1], input_dim_video[2]))
# This means it's indexable and needs at least 3 elements
# input_dim_text is used as: Input(shape=(input_dim_text[1],))
# This means it's indexable and needs at least 2 elements

input_dim_video = (2, 5, 4)  # (batch_size, timesteps, features) - shape tuple
input_dim_text = (2, 3)      # (batch_size, text_dim) - shape tuple

ARGS = {
    'input_dim_video': input_dim_video,
    'input_dim_text': input_dim_text,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
keras — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Functional — a Functional output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Functional 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
fleiss_kappa Ran
ran on a synthesized input
MichiganNLP/vlog_action_recognition/amt/compute_agreement.py
code served (permissive licence) · get_code("69de3cbbda85c060")
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 float, ndim 0, dtype family float, 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

torch.manual_seed(0)

# Construct ratings as a list of (item, category) tuples
# Let's create a scenario with 3 items and 4 raters rating them into 2 categories
ratings = [
    (0, 'A'),  # item 0, category A
    (0, 'A'),  # item 0, category A
    (0, 'B'),  # item 0, category B
    (0, 'A'),  # item 0, category A
    (1, 'A'),  # item 1, category A
    (1, 'A'),  # item 1, category A
    (1, 'A'),  # item 1, category A
    (1, 'B'),  # item 1, category B
    (2, 'B'),  # item 2, category B
    (2, 'B'),  # item 2, category B
    (2, 'B'),  # item 2, category B
    (2, 'A'),  # item 2, category A
]

# n is the number of raters (fixed number per item)
n = 4

ARGS = {
    'ratings': ratings,
    'n': n,
}
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 float, ndim 0, dtype family float, 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
get_compromised_hits Ran
ran on a synthesized input
MichiganNLP/vlog_action_recognition/amt/detect_spam.py
code served (permissive licence) · get_code("8aec1885fd917bd9")
How it ran

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

import torch

torch.manual_seed(0)

# Construct potential_spammers dict
# Keys are hit_nb (identifiers), values are lists of spammer worker IDs
potential_spammers = {
    "hit_001": ["worker_1", "worker_2"],  # len > 1, so this HIT is compromised
    "hit_002": ["worker_3", "worker_4", "worker_5"],  # len > 1, so this HIT is compromised
    "hit_003": ["worker_6"],  # len == 1, so this HIT is NOT compromised
    "hit_004": ["worker_7", "worker_8"],  # len > 1, so this HIT is compromised
}

ARGS = {
    "potential_spammers": potential_spammers
}
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 tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
read_results_from_AMT Ran
ran on a synthesized input
MichiganNLP/vlog_action_recognition/amt/read_csv_results.py
code served (permissive licence) · get_code("dad39214a8dcfe23")
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 csv
import tempfile
from collections import OrderedDict

# Create a temporary CSV file with the expected structure
temp_csv = tempfile.NamedTemporaryFile(mode='w', suffix='.csv', delete=False)

# Write CSV header and sample data
csv_writer = csv.writer(temp_csv)
csv_writer.writerow([
    'HITId',
    'Input.video_url1',
    'Input.video_url2',
    'Input.video_url3',
    'Input.video_url4',
    'Input.video_url5',
    'Input.actions1',
    'Input.actions2',
    'Input.actions3',
    'Input.actions4',
    'Input.actions5',
    'Answer.name1',
    'Answer.name2',
    'Answer.name3'
])

# First HIT
csv_writer.writerow([
    'HIT001',
    'video1.mp4',
    'video2.mp4',
    'video3.mp4',
    'video4.mp4',
    'video5.mp4',
    'action1;action2',
    'action3;action4',
    'action5;action6',
    'action7;action8',
    'action9;action10',
    'result_0_0_1',
    'result_0_1_2',
    ''
])

# Second HIT
csv_writer.writerow([
    'HIT002',
    'video6.mp4',
    'video7.mp4',
    'video8.mp4',
    'video9.mp4',
    'video10.mp4',
    'action11;action12',
    'action13;action14',
    'action15;action16',
    'action17;action18',
    'action19;action20',
    'result_0_0_3',
    'result_0_1_1',
    ''
])

temp_csv.close()

ARGS = {
    'csv_file_name': temp_csv.name
}
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 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
read_results_from_amt Ran
ran on a synthesized input
MichiganNLP/vlog_action_recognition/amt/read_AMT.py
code served (permissive licence) · get_code("872bab173b6ed8cc")
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 tempfile
import csv
from collections import OrderedDict

# Create a temporary CSV file with valid AMT results data
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.csv')
csv_file_path = temp_file.name

# Write header and sample data
writer = csv.DictWriter(
    temp_file,
    fieldnames=[
        'HITId',
        'Input.video_url1',
        'Input.video_url2',
        'Input.actions1',
        'Input.actions2',
        'Answer.name1',
        'Answer.name2',
    ]
)
writer.writeheader()

# Write first HIT data
writer.writerow({
    'HITId': 'HIT001',
    'Input.video_url1': 'http://example.com/video1.mp4',
    'Input.video_url2': 'http://example.com/video2.mp4',
    'Input.actions1': 'action1;action2',
    'Input.actions2': 'action3;action4',
    'Answer.name1': 'result_0_0_1',
    'Answer.name2': 'result_0_1_2',
})

# Write second HIT data
writer.writerow({
    'HITId': 'HIT002',
    'Input.video_url1': 'http://example.com/video3.mp4',
    'Input.video_url2': 'http://example.com/video4.mp4',
    'Input.actions1': 'action5;action6',
    'Input.actions2': 'action7;action8',
    'Answer.name1': 'result_0_0_2',
    'Answer.name2': 'result_0_1_3',
})

temp_file.close()

ARGS = {
    'csv_file_name': csv_file_path,
}
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 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
compare_with_GT Not yet run
unverified (harvested, no recorded run)
MichiganNLP/vlog_action_recognition/amt/detect_spam.py
code served (permissive licence) · get_code("3015fc5f426cd562")
How it ran

no invocation recorded.

getLength Not yet run
unverified (harvested, no recorded run)
MichiganNLP/vlog_action_recognition/amt/read_csv_results.py
code served (permissive licence) · get_code("8538771bd1dd8a7d")
How it ran

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, 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 11 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 consider the task of identifying human actions visible in online videos. We focus on the widely spread genre of lifestyle vlogs, which consist of videos of people performing actions while verbally describing them. Our goal is to identify if actions mentioned in the speech description of a video are visually present. We construct a dataset with crowdsourced manual annotations of visible actions, and introduce a multimodal algorithm that leverages information derived from visual and linguistic clues to automatically infer which actions are visible in a video. We demonstrate that our multimodal algorithm outperforms algorithms based only on one modality at a time.

For agents

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

get_harvested_code_for_paper("1906.04236")
get_code_for_paper("1906.04236")
have("1906.04236")

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

curl https://syntology.ai/api/ran/1906.04236.json

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

[![Syntology run record](https://syntology.ai/api/ran/1906.04236.svg)](https://syntology.ai/paper/1906.04236)

Connect an agent — have() is free.