SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1905.02822 · 2019

LightTrack: A Generic Framework for Online Top-Down Human Pose Tracking

arXiv · PDF · Open in the Atlas

Code that ran

11 of 16 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 10 have no contract verdict; 1 failure of our instrument; 5 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/1905.02822.json · badge: /api/ran/1905.02822.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
Guanghan/lighttrack canonical (classifier label)
classifier basis: census
9 ran · 4 no recorded run
NU-LL/lighttrack- reimplementation (classifier label)
classifier basis: classifier_2025
2 ran · 1 no recorded run
FunctionStatusWhere it lives
cal_BGR_for_img Ran
ran on a synthesized input
Guanghan/lighttrack/HPE/cal_posetrack_means.py
code served (permissive licence) · get_code("c5879c4e3c10dba1")
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

# Create a small test image with shape (height, width, 3) for BGR channels
# Using distinct dimensions to avoid ambiguity
height = 4
width = 5
img = np.random.rand(height, width, 3).astype(np.float32)

ARGS = {
    "img": img
}
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
change_order_COCO_to_PoseTrack Ran
ran on a synthesized input
Guanghan/lighttrack/HPE/AllJoints_COCO.py
code served (permissive licence) · get_code("6b3e1591b06619de")
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 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

# The function expects pose_keypoints_2d_COCO to be a list of floats
# with length that's a multiple of 3 (x, y, confidence for each keypoint)
# COCO has 17 keypoints, so we need 17 * 3 = 51 values

pose_keypoints_2d_COCO = [
    # Keypoint 0: nose
    100.0, 150.0, 0.9,
    # Keypoint 1: left_eye
    95.0, 145.0, 0.95,
    # Keypoint 2: right_eye
    105.0, 145.0, 0.95,
    # Keypoint 3: left_ear
    85.0, 140.0, 0.8,
    # Keypoint 4: right_ear
    115.0, 140.0, 0.8,
    # Keypoint 5: left_shoulder
    80.0, 200.0, 0.9,
    # Keypoint 6: right_shoulder
    120.0, 200.0, 0.9,
    # Keypoint 7: left_elbow
    70.0, 280.0, 0.85,
    # Keypoint 8: right_elbow
    130.0, 280.0, 0.85,
    # Keypoint 9: left_wrist
    60.0, 350.0, 0.8,
    # Keypoint 10: right_wrist
    140.0, 350.0, 0.8,
    # Keypoint 11: left_hip
    85.0, 350.0, 0.85,
    # Keypoint 12: right_hip
    115.0, 350.0, 0.85,
    # Keypoint 13: left_knee
    80.0, 450.0, 0.8,
    # Keypoint 14: right_knee
    120.0, 450.0, 0.8,
    # Keypoint 15: left_ankle
    75.0, 500.0, 0.75,
    # Keypoint 16: right_ankle
    125.0, 500.0, 0.75,
]

ARGS = {
    "pose_keypoints_2d_COCO": pose_keypoints_2d_COCO,
}
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 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
enlarge_bbox Ran
ran on a synthesized input
Guanghan/lighttrack/HPE/cal_posetrack_means.py
code served (permissive licence) · get_code("9d82c270a5e1fd34")
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)

# Analyze the function:
# - enlarge_bbox takes two parameters: bbox and scale
# - bbox is a tuple/list of 4 integers: [min_x, min_y, max_x, max_y]
# - scale is a positive number (float or int)
# - The function uses tuple unpacking: min_x, min_y, max_x, max_y = bbox
# - The function performs arithmetic operations on scale and bbox values
# - Returns a list of 4 integers

# Create valid arguments
bbox = [10, 20, 100, 150]  # [min_x, min_y, max_x, max_y]
scale = 0.5  # positive scale factor

ARGS = {
    "bbox": bbox,
    "scale": scale,
}
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
find_id_from_annotation_by_name Ran
ran on a synthesized input
Guanghan/lighttrack/jsonformat_std_to_posetrack18.py
code served (permissive licence) · get_code("6d66bdc0110d60fd")
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 torch
import numpy as np

def find(lst, key, value):
    index_list = []
    for i, dic in enumerate(lst):
        if dic[key] == value:
            index_list.append(i)
    return index_list

def find_id_from_annotation_by_name(gt_images_info, img_path):
    index_list = find(gt_images_info, key='file_name', value=img_path)
    assert len(index_list) >= 1
    index = index_list[0]
    frame_id = gt_images_info[index]['frame_id']
    return (frame_id, index)

# Construct gt_images_info as a list of dicts with 'file_name' and 'frame_id' keys
gt_images_info = [
    {'file_name': 'image_001.jpg', 'frame_id': 0},
    {'file_name': 'image_002.jpg', 'frame_id': 1},
    {'file_name': 'image_003.jpg', 'frame_id': 2},
    {'file_name': 'image_004.jpg', 'frame_id': 3},
]

# Set img_path to match one of the file_name values in gt_images_info
img_path = 'image_002.jpg'

ARGS = {
    'gt_images_info': gt_images_info,
    'img_path': img_path,
}
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 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
get_bbox_from_keypoints Ran
ran on a synthesized input
Guanghan/lighttrack/HPE/AllJoints_PoseTrack.py
code served (permissive licence) · get_code("f7e4031720659a65")
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

# Analyze the function get_bbox_from_keypoints
# It expects keypoints_python_data which is a list/sequence of keypoint objects
# Each keypoint object has 'x' and 'y' keys, where the values are indexable (accessed with [0])

# Create keypoint data structure
keypoints_python_data = [
    {'x': [10], 'y': [20]},
    {'x': [50], 'y': [30]},
    {'x': [40], 'y': [60]},
    {'x': [15], 'y': [45]},
]

ARGS = {
    'keypoints_python_data': keypoints_python_data
}
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
get_bbox_from_keypoints Ran
ran on a synthesized input
Guanghan/lighttrack/HPE/cal_posetrack_means.py
code served (permissive licence) · get_code("bfb81c803d0ab987")
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 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

torch.manual_seed(0)

# Construct valid arguments for get_bbox_from_keypoints
# The function expects keypoints_python_data which is a list/array of keypoint coordinates
# Format: [x0, y0, vis0, x1, y1, vis1, ...]
# where vis is visibility (0 = not visible, non-zero = visible)

keypoints_python_data = [
    10.0, 20.0, 1,      # keypoint 0: visible
    50.0, 60.0, 1,      # keypoint 1: visible
    100.0, 150.0, 1,    # keypoint 2: visible
    200.0, 250.0, 0,    # keypoint 3: not visible (will be ignored)
]

ARGS = {
    "keypoints_python_data": keypoints_python_data,
}
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, ndim 1, 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
get_joints_from_ann Ran
ran on a synthesized input
Guanghan/lighttrack/HPE/AllJoints_PoseTrack.py
code served (permissive licence) · get_code("c170cd33be80d235")
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)

# Construct a valid keypoints_python_data argument
# The function expects a list of dicts, where each dict has keys: 'id', 'x', 'y', 'is_visible'
# Each value is a list/array with at least one element (accessed via [0])

keypoints_python_data = [
    {
        'id': [0],
        'x': [100],
        'y': [200],
        'is_visible': [1]
    },
    {
        'id': [1],
        'x': [105],
        'y': [210],
        'is_visible': [0]
    },
    {
        'id': [2],
        'x': [110],
        'y': [220],
        'is_visible': [1]
    },
    {
        'id': [3],
        'x': [115],
        'y': [230],
        'is_visible': [1]
    },
    {
        'id': [5],
        'x': [120],
        'y': [240],
        'is_visible': [0]
    }
]

ARGS = {
    'keypoints_python_data': keypoints_python_data
}
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
get_pose_matching_score Ran
ran on a synthesized input
NU-LL/lighttrack-/demo_camera_mobile.py
code served (permissive licence) · get_code("3f1623059909000f")
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 float, 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 torch
import numpy as np
import sys

torch.manual_seed(0)

# Stand-in implementations for the helper functions called by get_pose_matching_score
def keypoints_to_graph(keypoints, bbox):
    """Convert keypoints and bbox to a graph representation"""
    # Return a simple graph object and a success flag
    graph = {
        'keypoints': keypoints,
        'bbox': bbox,
        'nodes': len(keypoints),
        'edges': []
    }
    flag_pass_check = True
    return graph, flag_pass_check

def graph_pair_to_data(sample_graph_pair):
    """Convert a pair of graphs to data objects"""
    graph_A, graph_B = sample_graph_pair
    # Return simple data objects
    data_A = {
        'graph': graph_A,
        'features': torch.randn(2, 4)
    }
    data_B = {
        'graph': graph_B,
        'features': torch.randn(2, 4)
    }
    return data_A, data_B

def pose_matching(data_A, data_B):
    """Match poses between two data objects"""
    # Return a match flag and distance
    flag_match = True
    dist = 0.5
    return flag_match, dist

# Valid test arguments
keypoints_A = [[100, 150], [200, 250], [300, 350]]
keypoints_B = [[110, 160], [210, 260], [310, 360]]
bbox_A = [50, 100, 400, 500]
bbox_B = [60, 110, 410, 510]

ARGS = {
    'keypoints_A': keypoints_A,
    'keypoints_B': keypoints_B,
    'bbox_A': bbox_A,
    'bbox_B': bbox_B
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 — no output shape is stated: observe_signature.py could not observe it (call:NameError)
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_track_id_SpatialConsistency Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
NU-LL/lighttrack-/demo_camera_mobile.py
code served (permissive licence) · get_code("71c1a19a1a7bc8c5")
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 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

def iou(boxA, boxB):
    xA = max(boxA[0], boxB[0])
    yA = max(boxA[1], boxB[1])
    xB = min(boxA[2], boxB[2])
    yB = min(boxA[3], boxB[3])
    interArea = max(0, xB - xA + 1) * max(0, yB - yA + 1)
    boxAArea = (boxA[2] - boxA[0] + 1) * (boxA[3] - boxA[1] + 1)
    boxBArea = (boxB[2] - boxB[0] + 1) * (boxB[3] - boxB[1] + 1)
    iou = interArea / float(boxAArea + boxBArea - interArea)
    return iou

def xywh_to_x1y1x2y2(det):
    x1, y1, w, h = det
    x2, y2 = (x1 + w, y1 + h)
    return [x1, y1, x2, y2]

def get_track_id_SpatialConsistency(bbox_cur_frame, bbox_list_prev_frame):
    thresh = 0.3
    max_iou_score = 0
    max_index = -1
    for bbox_index, bbox_det_dict in enumerate(bbox_list_prev_frame):
        bbox_prev_frame = bbox_det_dict['bbox']
        boxA = xywh_to_x1y1x2y2(bbox_cur_frame)
        boxB = xywh_to_x1y1x2y2(bbox_prev_frame)
        iou_score = iou(boxA, boxB)
        if iou_score > max_iou_score:
            max_iou_score = iou_score
            max_index = bbox_index
    if max_iou_score > thresh:
        track_id = bbox_list_prev_frame[max_index]['track_id']
        return (track_id, max_index)
    else:
        return (-1, None)

# Setup arguments
bbox_cur_frame = [10, 20, 50, 60]

bbox_list_prev_frame = [
    {
        'bbox': [12, 22, 48, 58],
        'track_id': 1
    },
    {
        'bbox': [100, 100, 150, 150],
        'track_id': 2
    }
]

ARGS = {
    'bbox_cur_frame': bbox_cur_frame,
    'bbox_list_prev_frame': bbox_list_prev_frame
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — 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
get_track_id_SpatialConsistency Ran
ran on a synthesized input
Guanghan/lighttrack/process_posetrack18_with_lighttrack_CPN101.py
code served (permissive licence) · get_code("04faef19f9d5d6ff")
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 int, ndim 0, dtype family int, 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.

# Setup code for get_track_id_SpatialConsistency

# Define bbox_gt: a ground truth bounding box in [x, y, w, h] format
bbox_gt = [10, 20, 50, 40]

# Define bbox_dets_list_list: a list of lists, where each inner list contains detection dictionaries
# Each detection dict has 'bbox' (in [x, y, w, h] format) and 'track_id'
bbox_dets_list_list = [
    # For img_id=1
    [
        {'bbox': [15, 25, 48, 38], 'track_id': 1},
        {'bbox': [100, 150, 50, 40], 'track_id': 2},
        {'bbox': [12, 22, 52, 42], 'track_id': 3},
    ],
    # For img_id=2
    [
        {'bbox': [5, 10, 60, 50], 'track_id': 4},
        {'bbox': [200, 200, 30, 30], 'track_id': 5},
    ]
]

# Define img_id: the image index (1-indexed)
img_id = 1

ARGS = {
    'bbox_gt': bbox_gt,
    'bbox_dets_list_list': bbox_dets_list_list,
    'img_id': img_id,
}
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 int, ndim 0, dtype family int, 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
x1y1x2y2_to_xywh Ran
ran on a synthesized input
Guanghan/lighttrack/HPE/AllJoints_PoseTrack.py
code served (permissive licence) · get_code("fb22a5f1796b7e0e")
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 x1y1x2y2_to_xywh takes a single parameter 'det'
# which is unpacked as x1, y1, x2, y2
# It expects an iterable with 4 elements that can be converted to int
# and supports indexing/unpacking

# Create a simple detection bounding box in x1, y1, x2, y2 format
det = [10, 20, 50, 80]

ARGS = {
    'det': det
}
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
get_pose_matching_score Not yet run
unverified (harvested, no recorded run)
Guanghan/lighttrack/process_posetrack18_with_lighttrack_CPN101.py
code served (permissive licence) · get_code("d4da552c9c8c3934")
How it ran

no invocation recorded.

get_track_id_SGCN Not yet run
unverified (harvested, no recorded run)
NU-LL/lighttrack-/demo_camera_mobile.py
code served (permissive licence) · get_code("0991530d60b0f3bb")
How it ran

no invocation recorded.

get_track_id_SGCN Not yet run
unverified (harvested, no recorded run)
Guanghan/lighttrack/process_posetrack18_with_lighttrack_CPN101.py
code served (permissive licence) · get_code("e0de7a00ba22b9b0")
How it ran

no invocation recorded.

standard_to_PoseTrack_17 Not yet run
unverified (harvested, no recorded run)
Guanghan/lighttrack/jsonformat_std_to_posetrack18.py
code served (permissive licence) · get_code("db64e8f955b8daec")
How it ran

no invocation recorded.

standard_to_PoseTrack_18 Not yet run
unverified (harvested, no recorded run)
Guanghan/lighttrack/jsonformat_std_to_posetrack18.py
code served (permissive licence) · get_code("e6da281eedfc4fb9")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 15 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 15 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

In this paper, we propose a novel effective light-weight framework, called LightTrack, for online human pose tracking. The proposed framework is designed to be generic for top-down pose tracking and is faster than existing online and offline methods. Single-person Pose Tracking (SPT) and Visual Object Tracking (VOT) are incorporated into one unified functioning entity, easily implemented by a replaceable single-person pose estimation module. Our framework unifies single-person pose tracking with multi-person identity association and sheds first light upon bridging keypoint tracking with object tracking. We also propose a Siamese Graph Convolution Network (SGCN) for human pose matching as a Re-ID module in our pose tracking system. In contrary to other Re-ID modules, we use a graphical representation of human joints for matching. The skeleton-based representation effectively captures human pose similarity and is computationally inexpensive. It is robust to sudden camera shift that introduces human drifting. To the best of our knowledge, this is the first paper to propose an online human pose tracking framework in a top-down fashion. The proposed framework is general enough to fit other pose estimators and candidate matching mechanisms. Our method outperforms other online methods while maintaining a much higher frame rate, and is very competitive with our offline state-of-the-art. We make the code publicly available at: https://github.com/Guanghan/lighttrack.

For agents

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

get_harvested_code_for_paper("1905.02822")
get_code_for_paper("1905.02822")
have("1905.02822")

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

curl https://syntology.ai/api/ran/1905.02822.json

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

[![Syntology run record](https://syntology.ai/api/ran/1905.02822.svg)](https://syntology.ai/paper/1905.02822)

Connect an agent — have() is free.