SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2209.09236 · ECCV · 2022

Real-time Online Video Detection with Temporal Smoothing Transformers

Yue Zhao, Philipp Krähenbühl

arXiv · PDF · Open in the Atlas

Code that ran

3 of 4 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 3 have no contract verdict; 0 failures of our instrument; 1 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/2209.09236.json · badge: /api/ran/2209.09236.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
zhaoyue-zephyrus/TeSTra canonical (classifier label)
classifier basis not recorded on the link
3 ran · 1 no recorded run
FunctionStatusWhere it lives
calibrated_average_precision_score Ran
ran on a synthesized input
zhaoyue-zephyrus/TeSTra/src/rekognition_online_action_detection/evaluation/metrics.py
code served (permissive licence) · get_code("4b0c45865a8c0eee")
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 float64, ndim 0, 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

# Construct test data for calibrated_average_precision_score
# The function expects:
# - y_true: array of binary labels (0 or 1)
# - y_score: array of prediction scores

np.random.seed(0)

# Create sample data
# y_true should be binary (0s and 1s)
y_true = np.array([1, 0, 1, 1, 0, 1, 0, 0])

# y_score should be prediction scores (floats between 0 and 1 typically)
y_score = np.array([0.9, 0.2, 0.8, 0.7, 0.3, 0.6, 0.4, 0.1])

ARGS = {
    "y_true": y_true,
    "y_score": y_score,
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, 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
get_stage_pred_scores Ran
ran on a synthesized input
zhaoyue-zephyrus/TeSTra/src/rekognition_online_action_detection/evaluation/metrics.py
code served (permissive licence) · get_code("2b4cae7268c1b0f8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct arguments for get_stage_pred_scores
# The function processes gt_targets (ground truth labels) and pred_scores (prediction scores)
# It segments data based on action boundaries and extracts specific stages

# Create a simple example with some actions (non-zero) and background (zero)
# Pattern: background, action1, background, action2, background
gt_targets = [0, 1, 1, 0, 2, 2, 2, 0]
pred_scores = [0.1, 0.8, 0.9, 0.2, 0.7, 0.85, 0.75, 0.15]

# Percentages for stage start and end within each action
perc_s = 0.3  # Start at 30% through the action
perc_e = 0.8  # End at 80% through the action

ARGS = {
    "gt_targets": gt_targets,
    "pred_scores": pred_scores,
    "perc_s": perc_s,
    "perc_e": perc_e,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
postprocessing Ran
ran on a synthesized input
zhaoyue-zephyrus/TeSTra/src/rekognition_online_action_detection/evaluation/postprocessing.py
code served (permissive licence) · get_code("a1406d77c11f45ee")
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 function. 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

# The postprocessing function takes data_name as parameter
# and returns a postprocessing function (thumos_postprocessing in this case)
# The returned function takes ground_truth, prediction, smooth, and switch parameters

# Create sample ground_truth and prediction arrays
# ground_truth needs column 21 to check validity
# prediction needs columns 5 and 8 for the switch operation
# Let's create arrays with at least 22 columns for ground_truth and 9 columns for prediction

np.random.seed(0)

# Create ground_truth: shape (10, 22) - 10 frames, 22 columns
# Column 21 contains the validity flag (0 or 1)
ground_truth = np.random.rand(10, 22)
ground_truth[:, 21] = np.array([0, 1, 0, 0, 1, 0, 0, 0, 1, 0])  # Mix of valid (0) and invalid (1)

# Create prediction: shape (10, 9) - 10 frames, 9 columns
# Columns 5 and 8 are used in the switch operation
prediction = np.random.rand(10, 9)

ARGS = {
    "data_name": "THUMOS"
}

# The postprocessing function returns a callable, which we would then call with:
# postprocessing_fn = postprocessing("THUMOS")
# result = postprocessing_fn(ground_truth, prediction, smooth=False, switch=False)
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
perframe_average_precision Not yet run
unverified (harvested, no recorded run)
zhaoyue-zephyrus/TeSTra/src/rekognition_online_action_detection/evaluation/metrics.py
code served (permissive licence) · get_code("d8d010da99929e02")
How it ran

no invocation recorded.

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 0 later) and 0 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
calibrated_average_precision_score 1 other paper
1 dated earlier
2107.03377 (2021-07)
Long Short-Term Transformer for Online Action Detection
get_stage_pred_scores 1 other paper
1 dated earlier
2107.03377 (2021-07)
Long Short-Term Transformer for Online Action Detection
perframe_average_precision 1 other paper
1 dated earlier
2107.03377 (2021-07)
Long Short-Term Transformer for Online Action Detection
postprocessing 1 other paper
1 dated earlier
2107.03377 (2021-07)
Long Short-Term Transformer for Online Action Detection

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

Streaming video recognition reasons about objects and their actions in every frame of a video. A good streaming recognition model captures both long-term dynamics and short-term changes of video. Unfortunately, in most existing methods, the computational complexity grows linearly or quadratically with the length of the considered dynamics. This issue is particularly pronounced in transformer-based architectures. To address this issue, we reformulate the cross-attention in a video transformer through the lens of kernel and apply two kinds of temporal smoothing kernel: A box kernel or a Laplace kernel. The resulting streaming attention reuses much of the computation from frame to frame, and only requires a constant time update each frame. Based on this idea, we build TeSTra, a Temporal Smoothing Transformer, that takes in arbitrarily long inputs with constant caching and computing overhead. Specifically, it runs 6× faster than equivalent sliding-window based transformers with 2,048 frames in a streaming setting. Furthermore, thanks to the increased temporal span, TeSTra achieves state-of-the-art results on THUMOS'14 and EPIC-Kitchen-100, two standard online action detection and action anticipation datasets. A real-time version of TeSTra outperforms all but one prior approaches on the THUMOS'14 dataset.

For agents

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

get_harvested_code_for_paper("2209.09236")
get_code_for_paper("2209.09236")
have("2209.09236")

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

curl https://syntology.ai/api/ran/2209.09236.json

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

[![Syntology run record](https://syntology.ai/api/ran/2209.09236.svg)](https://syntology.ai/paper/2209.09236)

Connect an agent — have() is free.