SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.14525 · 2020

MobileDets: Searching for Object Detection Architectures for Mobile Accelerators

arXiv · PDF · Open in the Atlas

Code that ran

9 of 10 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 0 failures of our instrument; 1 no recorded run), measured 2026-11-21; 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/2004.14525.json · badge: /api/ran/2004.14525.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
inacmor/mobiledets-yolov4-pytorch community (classifier label)
classifier basis not recorded on the link
7 ran · 1 no recorded run
DataCanvasIO/Hypernets community (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
FunctionStatusWhere it lives
avg_iou Ran
ran on a synthesized input
inacmor/mobiledets-yolov4-pytorch/kmeans.py
code served (permissive licence) · get_code("4c6b7d44826bea5a")
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

# Construct arguments for avg_iou function
# The function expects:
# - box: numpy array of shape (N, 2) containing box dimensions
# - cluster: numpy array of shape (M, 2) containing cluster dimensions

# Create small stand-in data
box = np.array([
    [10.0, 20.0],
    [15.0, 25.0]
], dtype=np.float32)

cluster = np.array([
    [12.0, 18.0],
    [8.0, 22.0],
    [14.0, 24.0]
], dtype=np.float32)

ARGS = {
    "box": box,
    "cluster": cluster
}
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
calc_nondominated_set Ran
ran on a synthesized input
DataCanvasIO/Hypernets/hypernets/core/pareto.py
code served (permissive licence) · get_code("170bab6b8bf13844")
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 numpy as np

# Create a small set of solutions for testing
# Each solution has 3 objectives (columns)
# We'll create 4 solutions (rows) with some that dominate others
solutions = np.array([
    [1.0, 2.0, 3.0],      # Solution 0: potentially pareto optimal
    [2.0, 1.0, 4.0],      # Solution 1: potentially pareto optimal
    [3.0, 3.0, 5.0],      # Solution 2: dominated by others
    [1.5, 2.5, 3.5],      # Solution 3: potentially pareto optimal
], dtype=np.float32)

# Define directions: 'min' means we want to minimize that objective
directions = ['min', 'min', 'min']

ARGS = {
    'solutions': solutions,
    'dominate_func': None,
    'directions': directions
}
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 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
cas_iou Ran
ran on a synthesized input
inacmor/mobiledets-yolov4-pytorch/kmeans.py
code served (permissive licence) · get_code("364d90eb3c67abdf")
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 (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

# Analyze the function:
# - box: used with indexing box[0] and box[1], appears to be a 1D array-like with at least 2 elements
# - cluster: used with indexing cluster[:, 0] and cluster[:, 1], appears to be a 2D array with shape (n, 2+)
# The function computes IoU (Intersection over Union) for bounding boxes
# box represents a single box with [width, height]
# cluster represents multiple boxes with shape (num_boxes, 2) where each row is [width, height]

# Create valid arguments
box = np.array([10.0, 8.0])  # Single box with width=10, height=8

cluster = np.array([
    [5.0, 4.0],   # First cluster box
    [12.0, 6.0],  # Second cluster box
    [8.0, 9.0],   # Third cluster box
])

ARGS = {
    "box": box,
    "cluster": cluster
}
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 (3,) — shape (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
convert Ran
ran on a synthesized input
inacmor/mobiledets-yolov4-pytorch/dataset/data_augmentation.py
code served (permissive licence) · get_code("804465e5dbfd6cf3")
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, 4) 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 random

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

# Create a small test image as a numpy array
# Using shape (2, 4) for a tiny 2x4 image with pixel values
img = np.array([[100.0, 150.0, 200.0, 50.0],
                [120.0, 180.0, 220.0, 80.0]])

# Define contrast parameter (must be >= 1 for random.uniform to work with it as upper bound)
contrast = 2.0

# Define brightness range parameters
bri_low = -10.0
bri_up = 10.0

ARGS = {
    "img": img,
    "contrast": contrast,
    "bri_low": bri_low,
    "bri_up": bri_up
}
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, 4) — shape (2, 4) 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
flip Ran
ran on a synthesized input
inacmor/mobiledets-yolov4-pytorch/dataset/data_augmentation.py
code served (permissive licence) · get_code("c3a70d39e1470439")
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 cv2
import numpy as np
import random

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

# Create a small test image (square, as the function uses img.shape[0] for dimension s)
img = np.random.randint(0, 256, (64, 64, 3), dtype=np.uint8)

# Create a small box array with shape (2, 5)
# Columns: [id, y1, x1, y2, x2] based on how the function indexes them
box = np.array([
    [0, 10, 15, 30, 40],
    [1, 20, 25, 50, 55]
], dtype=np.float32)

# Mode can be 'hori', 'vert', or 'mix'
mode = 'hori'

ARGS = {
    'img': img,
    'box': box,
    'mode': mode
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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
kmeans Ran
ran on a synthesized input
inacmor/mobiledets-yolov4-pytorch/kmeans.py
code served (permissive licence) · get_code("e93bee702f3648bb")
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

# Construct a small bounding box dataset
# Each row represents a bounding box with [width, height]
np.random.seed(42)
box = np.array([
    [10.0, 20.0],
    [12.0, 22.0],
    [50.0, 60.0],
    [52.0, 58.0],
    [100.0, 120.0],
    [102.0, 118.0]
], dtype=np.float32)

# Number of clusters
k = 3

ARGS = {
    "box": box,
    "k": k
}
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
pareto_dominate Ran
ran on a synthesized input
DataCanvasIO/Hypernets/hypernets/core/pareto.py
code served (permissive licence) · get_code("592bbcceb75b8e44")
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 bool, 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

# Test case 1: x1 dominates x2 in minimization
x1_min = np.array([1.0, 2.0, 3.0])
x2_min = np.array([2.0, 3.0, 4.0])
directions_min = ['min', 'min', 'min']

# Test case 2: x1 dominates x2 in maximization
x1_max = np.array([5.0, 6.0, 7.0])
x2_max = np.array([4.0, 5.0, 6.0])
directions_max = ['max', 'max', 'max']

# Test case 3: mixed directions
x1_mixed = np.array([1.0, 6.0])
x2_mixed = np.array([2.0, 5.0])
directions_mixed = ['min', 'max']

# Test case 4: default directions (None, should use 'min')
x1_default = [0.5, 1.5]
x2_default = [1.0, 2.0]

ARGS = {
    'x1': x1_min,
    'x2': x2_min,
    'directions': directions_min
}
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 bool, 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
rotate Ran
ran on a synthesized input
inacmor/mobiledets-yolov4-pytorch/dataset/data_augmentation.py
code served (permissive licence) · get_code("953567eb4c4700a9")
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
import random

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

# Create a small square image (s x s x 3 for RGB)
s = 8
img = np.random.randint(0, 256, size=(s, s, 3), dtype=np.uint8)

# Create a box array with shape (num_boxes, 5)
# Each box has 5 elements: [class_id, x1, y1, x2, y2]
# Using distinct dimensions for clarity
box = np.array([
    [0, 1.0, 2.0, 3.0, 4.0],
    [1, 2.0, 3.0, 5.0, 6.0]
], dtype=np.float32)

ARGS = {
    "img": img,
    "box": box
}
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
smooth_label Ran
ran on a synthesized input
inacmor/mobiledets-yolov4-pytorch/model/loss.py
code served (permissive licence) · get_code("de22edbfdf0d75e6")
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, 4) 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

torch.manual_seed(0)

# ground_truth should be a tensor with one-hot encoded labels or probability distributions
# The function accesses .size(-1) to get num_classes
# Shape: (batch_size, num_classes)
ground_truth = torch.tensor([
    [1.0, 0.0, 0.0, 0.0],
    [0.0, 1.0, 0.0, 0.0]
], dtype=torch.float32)

# label_smoothing should be a float between 0 and 1
label_smoothing = 0.1

ARGS = {
    "ground_truth": ground_truth,
    "label_smoothing": label_smoothing
}
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, 4) — shape (2, 4) 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
resize_image Not yet run
unverified (harvested, no recorded run)
inacmor/mobiledets-yolov4-pytorch/dataset/datasets.py
code served (permissive licence) · get_code("cfd33b179e7cf880")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 2 the same month, 0 later) and 8 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
avg_iou 1 other paper
1 the same month
2004.10934 (2020-04)
YOLOv4: Optimal Speed and Accuracy of Object Detection
cas_iou 1 other paper
1 the same month
2004.10934 (2020-04)
YOLOv4: Optimal Speed and Accuracy of Object 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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Inverted bottleneck layers, which are built upon depthwise convolutions, have been the predominant building blocks in state-of-the-art object detection models on mobile devices. In this work, we investigate the optimality of this design pattern over a broad range of mobile accelerators by revisiting the usefulness of regular convolutions. We discover that regular convolutions are a potent component to boost the latency-accuracy trade-off for object detection on accelerators, provided that they are placed strategically in the network via neural architecture search. By incorporating regular convolutions in the search space and directly optimizing the network architectures for object detection, we obtain a family of object detection models, MobileDets, that achieve state-of-the-art results across mobile accelerators. On the COCO object detection task, MobileDets outperform MobileNetV3+SSDLite by 1.7 mAP at comparable mobile CPU inference latencies. MobileDets also outperform MobileNetV2+SSDLite by 1.9 mAP on mobile CPUs, 3.7 mAP on Google EdgeTPU, 3.4 mAP on Qualcomm Hexagon DSP and 2.7 mAP on Nvidia Jetson GPU without increasing latency. Moreover, MobileDets are comparable with the state-of-the-art MnasFPN on mobile CPUs even without using the feature pyramid, and achieve better mAP scores on both EdgeTPUs and DSPs with up to 2x speedup. Code and models are available in the TensorFlow Object Detection API: https://github.com/tensorflow/models/tree/master/research/object_detection.

For agents

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

get_harvested_code_for_paper("2004.14525")
get_code_for_paper("2004.14525")
have("2004.14525")

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

curl https://syntology.ai/api/ran/2004.14525.json

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

[![Syntology run record](https://syntology.ai/api/ran/2004.14525.svg)](https://syntology.ai/paper/2004.14525)

Connect an agent — have() is free.