SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1902.04103 · 2019

Bag of Freebies for Training Object Detection Neural Networks

arXiv · PDF · Open in the Atlas

Code that ran

15 of 18 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 13 have no contract verdict; 2 failures of our instrument; 3 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/1902.04103.json · badge: /api/ran/1902.04103.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
wizyoung/YOLOv3_TensorFlow community (classifier label)
classifier basis not recorded on the link
15 ran · 3 no recorded run
FunctionStatusWhere it lives
avg_iou Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/get_kmeans.py
code served (permissive licence) · get_code("34e841aabf1b96f9")
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 float32, 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

# Create small sample data for testing avg_iou
# boxes: shape (r, 2) - r bounding boxes with width and height
boxes = np.array([
    [10.0, 20.0],
    [15.0, 25.0],
    [12.0, 18.0]
], dtype=np.float32)

# clusters: shape (k, 2) - k cluster centers with width and height
clusters = np.array([
    [10.0, 20.0],
    [14.0, 24.0]
], dtype=np.float32)

ARGS = {
    "boxes": boxes,
    "clusters": clusters
}
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 float32, 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
bbox_crop Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/data_aug.py
code served (permissive licence) · get_code("00fa5fa35fceddc1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (3, 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

# Create a sample bounding box array with shape (N, 4+)
# N=3 bounding boxes, each with 4 coordinates (x_min, y_min, x_max, y_max)
# plus one additional attribute
bbox = np.array([
    [10.0, 20.0, 50.0, 60.0, 1.0],
    [15.0, 25.0, 45.0, 55.0, 2.0],
    [5.0, 10.0, 30.0, 40.0, 3.0]
], dtype=np.float32)

# Define crop_box as (x_min, y_min, width, height)
crop_box = (0, 0, 100, 100)

# allow_outside_center controls whether to keep boxes with centers outside crop area
allow_outside_center = True

ARGS = {
    'bbox': bbox,
    'crop_box': crop_box,
    'allow_outside_center': allow_outside_center
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (3, 5) — shape (3, 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
bbox_iou Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/data_aug.py
code served (permissive licence) · get_code("bd08a00ded7caa44")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (3, 2) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create two sets of bounding boxes
# bbox_a: shape (3, 4) - 3 bounding boxes with [x1, y1, x2, y2] format
bbox_a = np.array([
    [0.0, 0.0, 10.0, 10.0],
    [5.0, 5.0, 15.0, 15.0],
    [20.0, 20.0, 30.0, 30.0]
], dtype=np.float32)

# bbox_b: shape (2, 4) - 2 bounding boxes with [x1, y1, x2, y2] format
bbox_b = np.array([
    [0.0, 0.0, 5.0, 5.0],
    [7.0, 7.0, 12.0, 12.0]
], dtype=np.float32)

# offset parameter
offset = 0

ARGS = {
    'bbox_a': bbox_a,
    'bbox_b': bbox_b,
    'offset': offset
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (3, 2) — shape (3, 2) 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
calc_iou Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/eval_utils.py
code served (permissive licence) · get_code("9db9b9965e68c8ad")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (2, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create small bounding boxes for testing
# Format: [x_min, y_min, x_max, y_max]
pred_boxes = np.array([
    [0.0, 0.0, 2.0, 2.0],
    [1.0, 1.0, 3.0, 3.0]
], dtype=np.float32)

true_boxes = np.array([
    [0.5, 0.5, 2.5, 2.5],
    [1.5, 1.5, 3.5, 3.5],
    [4.0, 4.0, 5.0, 5.0]
], dtype=np.float32)

ARGS = {
    "pred_boxes": pred_boxes,
    "true_boxes": true_boxes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (2, 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
cpu_nms Ran
ran; our contract draft was wrong, not the code
wizyoung/YOLOv3_TensorFlow/utils/nms_utils.py
code served (permissive licence) · get_code("f969b2977a72cc55")
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 numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create sample boxes: shape [1, 10647, 4]
# Each box has [x1, y1, x2, y2] coordinates
num_boxes = 10647
boxes = np.random.rand(1, num_boxes, 4) * 100
# Ensure x2 > x1 and y2 > y1
boxes[:, :, 2] = boxes[:, :, 0] + np.random.rand(1, num_boxes) * 50
boxes[:, :, 3] = boxes[:, :, 1] + np.random.rand(1, num_boxes) * 50

# Create sample scores: shape [1, 10647, num_classes]
num_classes = 5
scores = np.random.rand(1, num_boxes, num_classes)

ARGS = {
    "boxes": boxes,
    "scores": scores,
    "num_classes": num_classes,
    "max_boxes": 50,
    "score_thresh": 0.5,
    "iou_thresh": 0.5,
}
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 — 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
gpu_nms Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/nms_utils.py
code served (permissive licence) · get_code("5f0874e5158a0875")
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 tensorflow as tf
import numpy as np

# Set random seed for reproducibility
tf.random.set_seed(0)

# Define parameters
num_classes = 3
batch_size = 1
num_boxes = 100  # Small number of boxes for testing

# Create boxes tensor with shape [1, 100, 4]
# Boxes are in format [y1, x1, y2, x2] normalized to [0, 1]
boxes_data = np.random.rand(batch_size, num_boxes, 4).astype(np.float32)
# Ensure y1 < y2 and x1 < x2
boxes_data[:, :, [0, 2]] = np.sort(boxes_data[:, :, [0, 2]], axis=2)
boxes_data[:, :, [1, 3]] = np.sort(boxes_data[:, :, [1, 3]], axis=2)
boxes = tf.constant(boxes_data)

# Create scores tensor with shape [1, 100, 3]
# Scores are confidence * probability values between 0 and 1
scores_data = np.random.rand(batch_size, num_boxes, num_classes).astype(np.float32)
scores = tf.constant(scores_data)

# Define remaining parameters
max_boxes = 50
score_thresh = 0.5
nms_thresh = 0.5

ARGS = {
    'boxes': boxes,
    'scores': scores,
    'num_classes': num_classes,
    'max_boxes': max_boxes,
    'score_thresh': score_thresh,
    'nms_thresh': nms_thresh,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — 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
iou Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/get_kmeans.py
code served (permissive licence) · get_code("95c4cba76be0a2a1")
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 1, dtype family float, shape (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

# Construct arguments for the iou function

# box: tuple or array representing width and height (shifted to origin)
box = (10.0, 8.0)

# clusters: numpy array of shape (k, 2) where k is the number of clusters
# Using k=5 clusters with distinct width and height values
clusters = np.array([
    [5.0, 4.0],
    [8.0, 6.0],
    [12.0, 10.0],
    [3.0, 2.0],
    [15.0, 12.0]
], dtype=np.float32)

ARGS = {
    "box": box,
    "clusters": clusters
}
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 1, dtype family float, shape (5,) — shape (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
make_summary Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/misc_utils.py
code served (permissive licence) · get_code("ff97bfe78626bb8e")
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 Summary. 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 tensorflow as tf
from tensorflow.core.framework import summary_pb2

# Construct valid arguments for make_summary
ARGS = {
    "name": "loss",
    "val": 0.5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Summary — a Summary output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Summary 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
mix_up Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/data_aug.py
code served (permissive licence) · get_code("b3ec10e70300afb7")
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 numpy as np

# Create sample images in HWC format
img1 = np.random.randint(0, 256, size=(32, 48, 3), dtype='uint8')
img2 = np.random.randint(0, 256, size=(40, 56, 3), dtype='uint8')

# Create sample bounding boxes [N, 4] format: x_min, y_min, x_max, y_max
bbox1 = np.array([
    [10, 15, 30, 35],
    [5, 20, 25, 40]
], dtype='float32')

bbox2 = np.array([
    [12, 18, 45, 50],
], dtype='float32')

ARGS = {
    'img1': img1,
    'img2': img2,
    'bbox1': bbox1,
    'bbox2': bbox2
}
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 — 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
parse_anchors Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/misc_utils.py
code served (permissive licence) · get_code("44e65a34fe58f184")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (4, 2) 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 tempfile
import os

# Create a temporary file with anchor data
temp_dir = tempfile.mkdtemp()
anchor_file = os.path.join(temp_dir, "anchors.txt")

# Write anchor data in the expected format (comma-separated values that reshape to [N, 2])
anchor_data = "10.5,20.3,15.2,25.1,30.0,40.5,12.3,18.9"
with open(anchor_file, 'w') as f:
    f.write(anchor_data)

ARGS = {
    "anchor_path": anchor_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (4, 2) — shape (4, 2) 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
parse_line Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/data_utils.py
code served (permissive licence) · get_code("7a494e675e71f1f9")
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 numpy as np

# The function parse_line takes a single parameter: line
# which should be either a string or bytes representing a line from an annotation file
# Format: line_index img_path img_width img_height [label x_min y_min x_max y_max] ...

# Create a valid annotation line with:
# - line index: 0
# - image path: "path/to/image.jpg"
# - image width: 640
# - image height: 480
# - one bounding box: label=1, x_min=10.5, y_min=20.5, x_max=100.5, y_max=150.5

line_string = "0 path/to/image.jpg 640 480 1 10.5 20.5 100.5 150.5 2 200.0 300.0 400.0 450.0"

ARGS = {
    "line": line_string
}
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 — 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
process_box Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/data_utils.py
code served (permissive licence) · get_code("fc500933be7ea183")
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 numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Define arguments for process_box function
boxes = np.array([
    [10.0, 20.0, 50.0, 80.0, 1.0],
    [100.0, 150.0, 200.0, 250.0, 0.8]
], dtype=np.float32)

labels = np.array([0, 2], dtype=np.int64)

img_size = (416, 416)

class_num = 4

anchors = np.array([
    [10, 13],
    [16, 30],
    [33, 23],
    [30, 61],
    [62, 45],
    [59, 119],
    [116, 90],
    [156, 198],
    [373, 326]
], dtype=np.float32)

ARGS = {
    "boxes": boxes,
    "labels": labels,
    "img_size": img_size,
    "class_num": class_num,
    "anchors": anchors
}
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 — 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
py_nms Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
wizyoung/YOLOv3_TensorFlow/utils/nms_utils.py
code served (permissive licence) · get_code("647b8865c23ab104")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) 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 numpy as np

# Create sample bounding boxes in format [x1, y1, x2, y2]
# Using 6 boxes to test NMS functionality
boxes = np.array([
    [10.0, 10.0, 50.0, 50.0],   # box 0
    [12.0, 12.0, 52.0, 52.0],   # box 1 - overlaps with box 0
    [100.0, 100.0, 150.0, 150.0], # box 2 - far from others
    [15.0, 15.0, 55.0, 55.0],   # box 3 - overlaps with box 0 and 1
    [200.0, 200.0, 250.0, 250.0], # box 4 - far from others
    [105.0, 105.0, 155.0, 155.0]  # box 5 - overlaps with box 2
], dtype=np.float32)

# Create corresponding scores (higher score = more confident detection)
scores = np.array([0.9, 0.75, 0.85, 0.6, 0.95, 0.7], dtype=np.float32)

# Set max_boxes and iou_thresh parameters
max_boxes = 50
iou_thresh = 0.5

ARGS = {
    'boxes': boxes,
    'scores': scores,
    'max_boxes': max_boxes,
    'iou_thresh': iou_thresh
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy — 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
read_class_names Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/utils/misc_utils.py
code served (permissive licence) · get_code("9f9d523c4fc9a5d5")
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 dict. 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 tempfile
import os

# Create a temporary file with class names
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt')
temp_file.write("cat\n")
temp_file.write("dog\n")
temp_file.write("bird\n")
temp_file.write("fish\n")
temp_file.close()

class_name_path = temp_file.name

ARGS = {
    "class_name_path": class_name_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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
translate_boxes Ran
ran on a synthesized input
wizyoung/YOLOv3_TensorFlow/get_kmeans.py
code served (permissive licence) · get_code("b9b52f6b2d452deb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (2, 2) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a small test case with 2 boxes, each with 4 coordinates (x1, y1, x2, y2)
boxes = np.array([
    [10.0, 20.0, 30.0, 50.0],
    [5.0, 15.0, 25.0, 45.0]
], dtype=np.float32)

ARGS = {
    "boxes": boxes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (2, 2) — shape (2, 2) 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
evaluate_on_gpu Not yet run
unverified (harvested, no recorded run)
wizyoung/YOLOv3_TensorFlow/utils/eval_utils.py
code served (permissive licence) · get_code("a21ccdee37daac79")
How it ran

no invocation recorded.

parse_data Not yet run
unverified (harvested, no recorded run)
wizyoung/YOLOv3_TensorFlow/utils/data_utils.py
code served (permissive licence) · get_code("c6da8f222212852c")
How it ran

no invocation recorded.

parse_xml Not yet run
unverified (harvested, no recorded run)
wizyoung/YOLOv3_TensorFlow/misc/parse_voc_xml.py
code served (permissive licence) · get_code("a6fbd5e031368b6e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 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, 1 later) and 16 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
cpu_nms 1 other paper
1 dated earlier
1506.02640 (2015-06)
You Only Look Once: Unified, Real-Time Object Detection
py_nms 1 other paper
1 later
1904.07850 (2019-04)
Objects as Points

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

Training heuristics greatly improve various image classification model accuracies~\cite{he2018bag}. Object detection models, however, have more complex neural network structures and optimization targets. The training strategies and pipelines dramatically vary among different models. In this works, we explore training tweaks that apply to various models including Faster R-CNN and YOLOv3. These tweaks do not change the model architectures, therefore, the inference costs remain the same. Our empirical results demonstrate that, however, these freebies can improve up to 5% absolute precision compared to state-of-the-art baselines.

For agents

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

get_harvested_code_for_paper("1902.04103")
get_code_for_paper("1902.04103")
have("1902.04103")

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

curl https://syntology.ai/api/ran/1902.04103.json

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

[![Syntology run record](https://syntology.ai/api/ran/1902.04103.svg)](https://syntology.ai/paper/1902.04103)

Connect an agent — have() is free.