SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2003.02501 · 2020

Detecting Attended Visual Targets in Video

arXiv · PDF · Open in the Atlas

Code that ran

11 of 13 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 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/2003.02501.json · badge: /api/ran/2003.02501.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
ejcgt/attention-target-detection community (classifier label)
classifier basis not recorded on the link
11 ran · 2 no recorded run
FunctionStatusWhere it lives
LSTMCell Ran
ran on a synthesized input
ejcgt/attention-target-detection/lib/pytorch_convolutional_rnn/convolutional_rnn/functional.py
code served (permissive licence) · get_code("73becc3e41a71a86")
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 3, 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.functional as F

torch.manual_seed(0)

# LSTMCell parameters:
# - input: tensor of shape (batch_size, input_size)
# - hidden: tuple of (hx, cx) where each is (batch_size, hidden_size)
# - w_ih: weight matrix (4*hidden_size, input_size)
# - w_hh: weight matrix (4*hidden_size, hidden_size)
# - b_ih: bias vector (4*hidden_size,) or None
# - b_hh: bias vector (4*hidden_size,) or None
# - linear_func: function to apply, defaults to F.linear

batch_size = 2
input_size = 3
hidden_size = 4

# Create input tensor
input_tensor = torch.randn(batch_size, input_size)

# Create hidden state tuple (hx, cx)
hx = torch.randn(batch_size, hidden_size)
cx = torch.randn(batch_size, hidden_size)
hidden = (hx, cx)

# Create weight matrices
w_ih = torch.randn(4 * hidden_size, input_size)
w_hh = torch.randn(4 * hidden_size, hidden_size)

# Create bias vectors
b_ih = torch.randn(4 * hidden_size)
b_hh = torch.randn(4 * hidden_size)

# linear_func defaults to F.linear if None is passed
linear_func = None

ARGS = {
    'input': input_tensor,
    'hidden': hidden,
    'w_ih': w_ih,
    'w_hh': w_hh,
    'b_ih': b_ih,
    'b_hh': b_hh,
    'linear_func': linear_func
}
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, ndim 3, 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
RNNReLUCell Ran
ran on a synthesized input
ejcgt/attention-target-detection/lib/pytorch_convolutional_rnn/convolutional_rnn/functional.py
code served (permissive licence) · get_code("05467393784434c1")
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, 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
import torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

# Small tensor dimensions: batch=2, input_dim=4, hidden_dim=3
batch_size = 2
input_dim = 4
hidden_dim = 3

# Create input tensor (batch_size, input_dim)
input_tensor = torch.randn(batch_size, input_dim)

# Create hidden state tensor (batch_size, hidden_dim)
hidden_tensor = torch.randn(batch_size, hidden_dim)

# Weight matrix for input-to-hidden: (hidden_dim, input_dim)
w_ih = torch.randn(hidden_dim, input_dim)

# Weight matrix for hidden-to-hidden: (hidden_dim, hidden_dim)
w_hh = torch.randn(hidden_dim, hidden_dim)

# Bias for input-to-hidden: (hidden_dim,)
b_ih = torch.randn(hidden_dim)

# Bias for hidden-to-hidden: (hidden_dim,)
b_hh = torch.randn(hidden_dim)

# linear_func is optional, defaults to F.linear if None
linear_func = None

ARGS = {
    "input": input_tensor,
    "hidden": hidden_tensor,
    "w_ih": w_ih,
    "w_hh": w_hh,
    "b_ih": b_ih,
    "b_hh": b_hh,
    "linear_func": linear_func,
}
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, 3) — shape (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
RNNTanhCell Ran
ran on a synthesized input
ejcgt/attention-target-detection/lib/pytorch_convolutional_rnn/convolutional_rnn/functional.py
code served (permissive licence) · get_code("cd1e300e333d5db3")
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, 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
import torch.nn.functional as F

torch.manual_seed(0)

# Define small tensors for the function
batch_size = 2
input_dim = 4
hidden_dim = 3

# input: tensor of shape (batch_size, input_dim)
input_tensor = torch.randn(batch_size, input_dim)

# hidden: tensor of shape (batch_size, hidden_dim)
hidden_tensor = torch.randn(batch_size, hidden_dim)

# w_ih: weight matrix for input-to-hidden, shape (hidden_dim, input_dim)
w_ih = torch.randn(hidden_dim, input_dim)

# w_hh: weight matrix for hidden-to-hidden, shape (hidden_dim, hidden_dim)
w_hh = torch.randn(hidden_dim, hidden_dim)

# b_ih: bias for input-to-hidden, shape (hidden_dim,)
b_ih = torch.randn(hidden_dim)

# b_hh: bias for hidden-to-hidden, shape (hidden_dim,)
b_hh = torch.randn(hidden_dim)

# linear_func: defaults to None, which uses F.linear
linear_func = None

ARGS = {
    'input': input_tensor,
    'hidden': hidden_tensor,
    'w_ih': w_ih,
    'w_hh': w_hh,
    'b_ih': b_ih,
    'b_hh': b_hh,
    'linear_func': linear_func
}
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, 3) — shape (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
ap Ran
ran on a synthesized input
ejcgt/attention-target-detection/utils/evaluation.py
code served (permissive licence) · get_code("b9f23783a1dde069")
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
from sklearn.metrics import average_precision_score

torch.manual_seed(0)

# average_precision_score expects:
# - y_true: array-like of true binary labels
# - y_score: array-like of target scores (predicted probabilities or decision function values)
# Both should be 1D arrays of the same length

label = np.array([0, 1, 1, 0, 1])
pred = np.array([0.1, 0.9, 0.8, 0.2, 0.7])

ARGS = {
    "label": label,
    "pred": pred
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, 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
argmax_pts Ran
ran on a synthesized input
ejcgt/attention-target-detection/utils/evaluation.py
code served (permissive licence) · get_code("f00c194f6adc3f25")
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

# The function argmax_pts takes a heatmap (numpy array)
# and finds the argmax position, then returns (x, y) coordinates
# We need to create a small 2D heatmap numpy array

heatmap = np.random.rand(4, 3)  # Small 2D array with distinct dimensions
heatmap[2, 1] = 0.99  # Set a clear maximum value at a specific location

ARGS = {
    "heatmap": heatmap
}
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
auc Ran
ran on a synthesized input
ejcgt/attention-target-detection/utils/evaluation.py
code served (permissive licence) · get_code("4672566e034d749b")
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 numpy as np
import torch

torch.manual_seed(0)

# The function takes:
# - heatmap: something that can be reshaped (has .size attribute or is numpy array)
# - onehot_im: something that can be reshaped (has .size attribute or is numpy array)
# - is_im: boolean flag

# When is_im=True, both heatmap and onehot_im are reshaped using .size
# This works for numpy arrays and torch tensors
# roc_auc_score expects 1D arrays of binary labels and scores

# Create small sample data
# heatmap: predicted scores/heatmap (continuous values)
heatmap = np.random.rand(4, 3).astype(np.float32)

# onehot_im: ground truth binary labels (reshaped to same size as heatmap)
onehot_im = np.random.randint(0, 2, size=(4, 3)).astype(np.float32)

# is_im: boolean flag indicating image mode
is_im = True

ARGS = {
    "heatmap": heatmap,
    "onehot_im": onehot_im,
    "is_im": is_im,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, 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_head_box_channel Ran
ran on a synthesized input
ejcgt/attention-target-detection/utils/imutils.py
code served (permissive licence) · get_code("e51e72ca0654a9a0")
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 (64, 64) 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)

ARGS = {
    'x_min': 10,
    'y_min': 20,
    'x_max': 100,
    'y_max': 150,
    'width': 200,
    'height': 300,
    'resolution': 64,
    'coordconv': False
}
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 Tensor, ndim 2, dtype family float, shape (64, 64) — shape (64, 64) 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
to_numpy Ran
ran, honoured the contract we drafted
ejcgt/attention-target-detection/utils/misc.py
pointer only (licence: MIT); the stored copy get_code serves, from SY-Xuan/IICS, is pointer only (licence: NONE) · get_code("b476623700945834")
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 SY-Xuan/IICS, 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.

to_torch Ran
ran, honoured the contract we drafted
ejcgt/attention-target-detection/utils/misc.py
pointer only (licence: MIT); the stored copy get_code serves, from liu-xb/GCMT, is pointer only (licence: NONE) · get_code("7aac4c51e0ceee97")
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 liu-xb/GCMT, 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.

unnorm Ran
ran on a synthesized input
ejcgt/attention-target-detection/utils/imutils.py
code served (permissive licence) · get_code("b65c0249984c6b66")
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 (3, 4, 5) 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

torch_seed = 0
np.random.seed(torch_seed)

# Create a small normalized image tensor (3 channels, 4x5 spatial dims)
img = np.random.randn(3, 4, 5).astype(np.float32)

# Define default parameters
mean = [0.485, 0.456, 0.406]
std = [0.229, 0.224, 0.225]

ARGS = {
    "img": img,
    "mean": mean,
    "std": std,
}
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 (3, 4, 5) — shape (3, 4, 5) 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
video_pack_sequences Ran
ran on a synthesized input
ejcgt/attention-target-detection/eval_on_videoatttarget.py
code served (permissive licence) · get_code("ebd6363b1de50a7e")
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 torch

torch.manual_seed(0)

# The function expects in_batch to be a list of tuples/sequences
# Each element in in_batch is a tuple where:
# - First element (index 0) is a tensor with shape (sequence_length, ...)
# - Additional elements are returned as-is in the output
# The function will pad all sequences to the max length

# Create sample sequences with variable lengths
seq1 = (torch.randn(5, 4), torch.tensor([1, 2, 3]))  # 5 timesteps, 4 features
seq2 = (torch.randn(3, 4), torch.tensor([4, 5, 6]))  # 3 timesteps, 4 features
seq3 = (torch.randn(4, 4), torch.tensor([7, 8, 9]))  # 4 timesteps, 4 features

in_batch = [seq1, seq2, seq3]

ARGS = {
    "in_batch": in_batch
}
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 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
draw_labelmap Not yet run
unverified (harvested, no recorded run)
ejcgt/attention-target-detection/utils/imutils.py
code served (permissive licence) · get_code("e15e27f2915b4002")
How it ran

no invocation recorded.

smooth_by_conv Not yet run
unverified (harvested, no recorded run)
ejcgt/attention-target-detection/utils/myutils.py
code served (permissive licence) · get_code("94dd3d9e6c6d6473")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 2 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 2 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.

FunctionAlso present inEarliest-datedOthers named
to_torch 9 other papers
1 dated earlier, 8 later · 3 only as an assembly brick
1912.08265 (2019-12)
Learning from Synthetic Animals
2007.09919 (2020-07)
Robust Tracking against Adversarial Attacks
2103.11658 (2021-03)
Intra-Inter Camera Similarity for Unsupervised Person Re-Identification
and 6 more not named here
to_numpy 5 other papers
5 later
2007.09919 (2020-07)
Robust Tracking against Adversarial Attacks
2103.11658 (2021-03)
Intra-Inter Camera Similarity for Unsupervised Person Re-Identification
136670327 (2022 (year only))
Spatial-Separated Curve Rendering Network for Efficient and High-Resolution Image Harmonization
and 2 more not named here

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 address the problem of detecting attention targets in video. Our goal is to identify where each person in each frame of a video is looking, and correctly handle the case where the gaze target is out-of-frame. Our novel architecture models the dynamic interaction between the scene and head features and infers time-varying attention targets. We introduce a new annotated dataset, VideoAttentionTarget, containing complex and dynamic patterns of real-world gaze behavior. Our experiments show that our model can effectively infer dynamic attention in videos. In addition, we apply our predicted attention maps to two social gaze behavior recognition tasks, and show that the resulting classifiers significantly outperform existing methods. We achieve state-of-the-art performance on three datasets: GazeFollow (static images), VideoAttentionTarget (videos), and VideoCoAtt (videos), and obtain the first results for automatically classifying clinically-relevant gaze behavior without wearable cameras or eye trackers.

For agents

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

get_harvested_code_for_paper("2003.02501")
get_code_for_paper("2003.02501")
have("2003.02501")

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

curl https://syntology.ai/api/ran/2003.02501.json

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

[![Syntology run record](https://syntology.ai/api/ran/2003.02501.svg)](https://syntology.ai/paper/2003.02501)

Connect an agent — have() is free.