drones

Article
Drone-YOLO: An Efficient Neural Network Method for Target
Detection in Drone Images
Zhengxin Zhang

College of Information Science and Technology, Zhongkai University of Agriculture and Engineering,
Guangzhou 510225, China; [email protected]

Abstract: Object detection in unmanned aerial vehicle (UAV) imagery is a meaningful foundation in
various research domains. However, UAV imagery poses unique challenges, including large image
sizes, small sizes detection objects, dense distribution, overlapping instances, and insufficient lighting
impacting the effectiveness of object detection. In this article, we propose Drone-YOLO, a series
of multi-scale UAV image object detection algorithms based on the YOLOv8 model, designed to
overcome the specific challenges associated with UAV image object detection. To address the issues
of large scene sizes and small detection objects, we introduce improvements to the neck component
of the YOLOv8 model. Specifically, we employ a three-layer PAFPN structure and incorporate a
detection head tailored for small-sized objects using large-scale feature maps, significantly enhancing
the algorithm’s capability to detect small-sized targets. Furthermore, we integrate the sandwich-
fusion module into each layer of the neck’s up–down branch. This fusion mechanism combines
network features with low-level features, providing rich spatial information about the objects at
different layer detection heads. We achieve this fusion using depthwise separable evolution, which
balances parameter costs and a large receptive field. In the network backbone, we employ RepVGG
modules as downsampling layers, enhancing the network’s ability to learn multi-scale features
and outperforming traditional convolutional layers. The proposed Drone-YOLO methods have
been evaluated in ablation experiments and compared with other state-of-the-art approaches on
the VisDrone2019 dataset. The results demonstrate that our Drone-YOLO (large) outperforms other
baseline methods in the accuracy of object detection. Compared to YOLOv8, our method achieves a
significant improvement in mAP0.5 metrics, with a 13.4% increase on the VisDrone2019-test and a
17.40% increase on the VisDrone2019-val. Additionally, the parameter-efficient Drone-YOLO (tiny)
Citation: Zhang, Z. Drone-YOLO: An
Efficient Neural Network Method for with only 5.25 M parameters performs equivalently or better than the baseline method with 9.66M
Target Detection in Drone Images. parameters on the dataset. These experiments validate the effectiveness of the Drone-YOLO methods
Drones 2023, 7, 526. https://doi.org/ in the task of object detection in drone imagery.
10.3390/drones7080526
Keywords: object detection; drone image; remote sensing
Academic Editor: Anastasios
Dimou

Received: 27 June 2023

Revised: 31 July 2023 1. Introduction
Accepted: 4 August 2023
In the past 15 years, with the gradual maturation of drone control technology, UAV
Published: 11 August 2023
remote sensing imagery has become an important data source in the field of low-altitude
remote sensing research due to its cost-effectiveness and ease of acquisition. During this
period, deep neural network methods have been extensively researched and have grad-
Copyright: © 2023 by the author.
ually become the optimal approaches for tasks such as image classification [1–3], object
Licensee MDPI, Basel, Switzerland. detection [4–6] and image segmentation [7–9]. However, the majority of currently applied
This article is an open access article deep neural network models, such as VGG [1], RESNET [2], U-NET [7], PSPNET [8], have
distributed under the terms and primarily been developed and validated using manually collected image datasets, such as
conditions of the Creative Commons VOC2007 [10],VOC2012 [11], MS-COCO [12], as shown in Figure 1.
Attribution (CC BY) license (https:// These images are collected based on the photographer’s subjective preferences.
creativecommons.org/licenses/by/ As shown in Figure 1: (1) many images are captured in small spaces scenes; (2) the number
4.0/). of objects in each image is relatively small, and the photographer will filter or avoid the

Drones 2023, 7, 526. https://doi.org/10.3390/drones7080526 https://www.mdpi.com/journal/drones

Drones 2023, 7, 526 2 of 23

objects or backgrounds, which are irrelevant with the main object; (3) there are not too
many densely arranged and overlapping targets in the image; (4) the image size is relatively
small; (5) the ratio of the size of the target in the image to the size of the entire image is
relatively large; (6) the image has good lighting conditions and no significant overexposure
or insufficient light; (7) many of the image-shooting angles are heads-up, and there are few
top-view images.

Figure 1. Sample images taken manually by people on the ground: (a) of a bicycle on a small road;
(b) Ponies in the stable; (c) A car and a man on the roadside; (d) A boat in the river; (e) Objects in the
bedroom; (f) Objects in the living room.

As shown in Figure 2, the images obtained from drones exhibit significant differ-
ences compared to manually captured ground-based imagery. These images taken by
drones have the following characteristics: (1) most of the images are outdoor scenes;
(2) the shooting angles of the images are mostly overhead views, with few upward per-
spectives; (3) the image scene is relatively large (outdoor images taken from at least
5 m above the ground); (4) there are many differently sized objects in the image. In many
scenes, objects are densely arranged. In some scenes, many similar-looking objects are
densely arranged and overlapped; (5) the image’s background is complex; (6) the sizes
of many images are large; (7) the ratio of the size of a single object to the size of the en-
tire image is relatively small; (8) some images have illumination problems, such as glare,
overexposure, and insufficient light.
In addition to these image data characteristics, there are two common application
scenarios for drone remote sensing object detection methods. The first one involves post-
flight data processing using large desktop computers. After the drone’s flight, the captured
data are processed on a desktop computer. The second one involves real-time processing
during the flight, where an embedded computer on the drone synchronously processes
the aerial imagery data in real time. This application is often used for obstacle avoidance
and automated mission planning during drone flights. Therefore, the applied neural
network object detection methods need to meet different requirements for each scenario.
For methods applicable to desktop computer environments, high detection accuracy is
required. For methods applicable to embedded environments, the model parameters need
to be within a certain scale to meet the embedded hardware operational requirements.
After meeting the operational conditions, the detection accuracy of the method also needs
to be as high as possible.
Drones 2023, 7, 526 3 of 23

Figure 2. Sample images taken from drones: (a) Crowds on the square; (b) The traffic flow on the
streets at night; (c) Setting up stalls and vehicles on daytime roads; (d) Buildings and docked ships
along the coast;.

Therefore, neural network methods for object detection in drone remote sensing
imagery need to be able to adapt to the specific characteristics of these data. They should
be designed to meet the requirements of post-flight data processing, which can provide
high accuracy and recall rates result, or they should be designed as models with smaller
scale parameters that can be deployed in embedded hardware environments for real-time
processing on drones.
In this work, we propose a series of improved neural network models based on the
YOLOv8 architecture with unified improvements. Among them, the models with larger
parameters can fulfill the requirements of post-flight data processing on desktop computers
in terms of object detection accuracy and recall rate. On the other hand, the models with
smaller scale parameters can meet the operational requirements of embedded hardware
environments and achieve satisfactory detection accuracy.
The main contributions of the article include:
1. A series of novel neural network methods for multi-scale object detection in UAV
images. These methods include high-accuracy models with larger parameters suitable
for desktop platforms and smaller parameter models suitable for deployment on
embedded edge computing devices. This provides a diverse range of effective options
for different computing environments.
2. A sandwich-fusion module with a large receptive field and low model parameters,
specifically designed for deployment in the neck network of the PAFPN structure.
Spatial features from the lower layers of the backbone network were effectively
captured by this module, leading to improved object detection performance.
3. The utilization of the RepVGG reparameterized convolution module to enhance the
downsampling layers in the network backbone. This modification improved the effective-
ness of the downsampling process, resulting in enhanced target detection performance.
Through ablation experiments, we demonstrate the feasibility and effectiveness of our
proposed network optimization methods. Experimental results on the VisDrone2019 dataset
show that our methods can improve the performances of the YOLOv8 series methods (large,
medium, small, tiny, and nano) under limited net model parameter growth. Addition-
ally, comparative analyses with state-of-the-art detection models and current mainstream
models with adjacent parameters demonstrate the superiority of our proposed method.
Drones 2023, 7, 526 4 of 23

The remaining sections of this paper are organized as follows: Section 2 provides a
review of the related work, and Section 3 elaborates on our proposed method. Section 4
presents the experimental results, and the conclusions are presented in Section 5.

2. Related Work
In recent years, numerous object detection neural network methods, particularly one-
stage object detection methods, have been proposed. Unlike the two-stage methods, the one-
stage method combines object location detection and classification in one step, achieving
real-time object detection on both desktop and embedded hardware. Among these works,
the YOLO series [13–20] received much attention. In particular, since YOLOv4 [15] was
proposed, many related works have been proposed. These methods not only achieve good
detection results, but also offer a series of improvements in areas such as network model
structures, activation functions, loss functions, network training methods, and training
data augmentation methods.
In the research on densely distributed object detection, Chu et al. [21] proposed a
neural network method, CrowdDet, for detecting dense and mutually occluded targets
in images. The authors believe that the primary reasons for poor performance in dense
object detection are (1) the presence of highly coincident instances (and their associated
candidate boxes) in the image space that are likely to have very similar features, making
it challenging for the detector to generate differentiated prediction results for each can-
didate box; (2) the severe overlap between instances, leading to prediction results being
suppressed by non-maximum suppression (NMS) errors. To solve the above problem, the
researchers proposed predicting a group of possibly highly coincident instance sets for
each candidate box of the detection target, instead of predicting a single instance as usual.
They introduced several methods: (1) EMD loss function, which is used to supervise the
learning process of instance set prediction; (2) the set NMS post-processing method to
suppress duplicate detection results of different candidate boxes and overcome the appli-
cation defects of NMS method in dense scenes; (3) the optimization module for handling
potential false positive targets. Bochkovskiy et al. [15] proposed YOLOv4. The author
introduced the CSPDarkNet [22] as the backbone structure of the network, enhancing the
learning abilities of convolution neural networks, allowing the network to maintain the
accuracy of feature map extraction while being lightweight; reducing computing bottle-
necks; and lowering memory costs. In the neck section of the network, the SPPF module
was introduced and the PAFPN module [23] was adopted. Jocher et al. [16] proposed a
one-stage target detection neural network model, YOLOv5. The method has high accuracy,
low computational complexity, and high speed. In the beginning structure of the backbone,
the author introduced the focus module. Compared with the convolutional layer with
3 × 3 kernel, this module has a higher computing speed, which mainly plays a role in
speeding up the processing of the input image. The neck section adopts a PAFPN module.
In the detection head section, an anchor-based method is deployed, and NMS is used to
eliminate erroneous candidate boxes. Ge et al. [17] proposed YOLOX, a one-stage, object
detection neural network model. The author still uses CSPDarkNet [22] as the backbone
network, but introduces SiLU [24] as the activation function, which solves the problem of
gradient dispersion when the input of the ReLU [25] function is negative and the output is
0. The author introduced decoupled heads, which separately implement confidence and
regression boxes, and integrate them into one during prediction; the author introduced
SimOTA to implement the anchor-free detector head. Li et al [18] proposed YOLOv6;
EfficientRep [26] was introduced as the backbone of the network. Li et al. [27] proposed
YOLOv6 3.0, which introduced the bidirectional concatenation (BiC) as the feature-map
fusion module deployed in the PAFPN structure of the network’s neck. Wang et al. [28] pro-
posed ELAN, a network module designed by controlling the shortest and longest gradient
paths. Wang at al. [19] proposed YOLOv7. The authors introduced an E-ELAN structure
in the backbone of this work, which can train multi-scale features without increasing the
Drones 2023, 7, 526 5 of 23

length of the shortest gradient path. Jocher et al. [20] proposed YOLOv8. The authors
introduced the C2f module as the backbone stage module.
Mao et al. [29] proposed MSA-YOLO, an object detection method. The authors made
improvements to the YOLOv5 model by introducing a multi-scale feature extraction module
(MFEM), which combined the squeeze-and-excitation module (SEM) and spatial attention
module (SAM) to form a hybrid attention mechanism known as the multi-scale split atten-
tion unit, serving as the backbone structure. In experiments compared with other network
models on the VisDrone2019-test dataset, the authors achieved the best performance in
terms of the mAP0.5 metric. Among all 10 types of detected objects, MSA-YOLO obtained
the best results in three types. Li et al. [30] proposed ACAM-YOLO, a modified version of
YOLOv5 [16], with adaptive co-attention module (ACAM) mechanisms. Through experi-
ments, the authors compared the difference in object detection results between the YOLOv5
network and the improved YOLOv5 using the ACAM. The experiment showed that the
network’s performance significantly improved, and the model parameters significantly
reduced. Zhao et al. [31] proposed MS-YOLOv7. This network is improved based on
YOLOv7 [19], increasing the original three detection heads to four, using the Swin trans-
former [32], W-MSA, SW-MSA, and the CBAM [33] attention mechanism to enhance the
features of the network neck. Using the soft-NMS method improves the performance of the
NMS method in densely distributed object detection. Li et al. [34] proposed an improved
small parameter object detection network based on the YOLOv8-s model. The authors
replaced the PAFPN structure of YOLOv8 with Bi-FPN [35] and improved the backbone
using the Ghostblock [36] module, achieving a neural network method with fewer pa-
rameters but better detection performance. Liu et al. [37] proposed EdgeYOLO, which
introduces a lite-decoupled head for object detection. In comparison with the decoupled
head of YOLOX [17], the detection head with fewer parameters achieves a faster inference
speed. Furthermore, it achieves comparable or superior accuracy compared to YOLOX.
Currently, one of the most significant challenges when applying these methods to
UAV object detection is in effectively detecting small objects. UAV remote sensing images
often have large image sizes, complex backgrounds, and a substantial presence of small
objects. Our proposed solution focuses on optimizing the accurate detection of small objects.
The advantage of YOLO series networks lies in their utilization of multi-level detection
heads, enabling the detection of objects of various sizes from different levels of feature
vectors. Our approach primarily revolves around the detection of tiny and small-sized
objects using feature vectors from lower layers, which have a higher spatial resolution.
To achieve this goal, we employ the sandwich-fusion module to optimize their semantic
features. The detection heads can obtain feature vectors with both high spatial resolution
and accurate semantic information, thereby enhancing the overall detection performance.

3. Methodology
Figure 3 shows the architecture of our proposed Drone-YOLO (large) network model.
This network structure is an improvement on the YOLOv8-l model. In the backbone section
of the network, we utilize the RepVGG structural re-parameterized convolution module
as the downsampling layer. During training, this convolutional structure simultaneously
trains both 3 × 3 and 1 × 1 convolutions. The two convolution kernels are merged into one
3 × 3 convolution layer during inference. This mechanism enables the network to learn
more robust features without compromising inference speed or inflating the model size. In
the neck section, we extend the PAFPN structure to three layers and append a small-size
object detection head. By incorporating the proposed sandwich-fusion module, spatial
and channel features are extracted from three different layer feature maps of the network’s
backbone. This optimization enhances the multi-scale detection head’s ability to gather
spatial positioning information of the objects to be detected.
We further refine the parameter size and model structure of the Drone-YOLO (large)
network model to obtain the Drone-YOLO series of networks. Among them, the Drone-
YOLO (tiny) and Drone-YOLO (nano) models have fewer parameters, yet achieve compa-
Drones 2023, 7, 526 6 of 23

rable detection accuracy to recently proposed models. This network model series achieved
good detection results on the VisDrone2019 dataset without utilizing widely employed
attention mechanisms.

Figure 3. Proposed network model (large) with its backbone, neck and head.

3.1. Backbone
The backbone of our proposed network model includes a stem layer and four stage
blocks. As shown in Figure 3, the stem layer is a convolutional layer with a stride of 2. Each
stage block has a RepVGG block with a stride of 2, and is used as a downsampling module
and a C2f convolutional block component, as proposed in YOLOv8.
In the proposed network, the feature map extraction backbone’s essential multi-stage
components are the C2f modules, which were modified from the C3 modules. In the
C3 module of CSPDarkNet, the backbone of YOLOv5, the cross-stage partial structure
was implemented and obtained good results. Its design goal is to reduce the network’s
computational complexity and enhance gradient performance. The design concept of
DenseNet [3] inspired the C3 module’s structure. The C3 module is composed of three
convolutional layers with a 1 × 1 kernel and a DarkNet bottleneck, which has a sequence
of convolutional layers with a 3 × 3 kernel. The module’s input feature map is divided
into two parts: One does not pass through the bottleneck, and the other passes through the
bottleneck. In the part of the feature map used to compute through the bottleneck, the first
layer’s output result is concatenated with its input as the input of the second convolutional
layer. The feature map passed through the bottleneck concatenates the other part feature
map, which only computes through one 1 × 1 convolution layer and passes through a final
1 × 1 convolutional layer to be the output feature of the C3 module.
In the backbone of this proposed network, we use the C2f structure from the YOLOv8
network as each stage unit of the backbone. As shown in Figure 4, compared with the C3
structure, the C2f module sufficiently meets the idea proposed by the ELAN module [28],
Drones 2023, 7, 526 7 of 23

which optimizes the network structure from the perspective of controlling the shortest
and longest gradient paths, thereby making the network more trainable. In addition,
without changing the network’s shortest and longest gradient paths, the C2f module can
effectively learn multi-scale features and expand the range of receptive fields through
feature vector diversion and multi-level-nested convolution within its module.

Figure 4. C2f convolutional block structure of YOLOv8.

We deployed RepVGG [38] blocks as the downsampling structure between stages in

the backbone network. The RepVGG component is improved from the VGG [1], which
is an early proposed single path network characterized by the extensive use of 3 × 3 con-
volutions as its essential units. The VGG is a memory-saving model because it does not
require memory for identity structures like residual blocks in ResNet [2]. The 3 × 3 convolu-
tional layer is a high-efficiency network structure. Machine learning hardware acceleration
libraries, such as cudnn [39] and MKL, have significantly optimized the computational
efficiency of 3 × 3 convolution, making the calculation density (theoretical FLOPs/time
usage) [38] of 3 × 3 convolution reach four times that of 1 × 1 or 5 × 5 convolution. Conse-
quently, extensive utilization of this type of convolutional neural network offers efficiency
advantages in practical computing tasks.
As shown in Figure 5, in the RepVGG module, there are two different convolutional
kernels, 3 × 3 and 1 × 1, during the training phase. This is due to the consistent movement
process of convolutional kernels of different sizes when calculated on feature maps. When
the model is used for inference, the 1 × 1 and 3 × 3 convolution kernels can be combined
into a single 3 × 3 kernel through structural re-parameterization. The specific approach
involves padding the surrounding part of the 1 × 1 kernel into a 3 × 3 form. Based on the
additivity principle of convolution kernels of the same size, the padded kernel is added to
the original 3 × 3 convolution kernel to form a 3 × 3 convolution kernel for inference.

Figure 5. RepVGG layer used as stage downsampling: (a) RepVGG layer in training; (b) RepVGG
layer in inference.
Drones 2023, 7, 526 8 of 23

3.2. Neck
In the proposed network, as shown in Figure 6, the neck’s structure is a PAFPN
structure network. The neck consists of top-down and bottom-up branches. In the top-
down branch, we have top-down layer 1, layer 2, and layer3. The different layers in the
top-down branch receive feature maps P1, P2, P3, P4, and P5 from the backbone’s stem
layer, and stage layer 1, stage layer 2, stage layer 3, and stage layer 4 through the SPPF
modules, respectively. The bottom-up branch is composed of three parts: layer 0, layer 1,
and layer 2. The input of the bottom-up comes from the output of the top-down branch,
as well as the feature map of the backbone’s stage layer 4 through SPPF. Their output
consists of four feature maps of different sizes, C2, C3, C4, and C5, corresponding to the
detection heads of targets of different sizes.

Figure 6. The neck of Drone-YOLO (large).

Given that the image data size we need to analyze is relatively large and the objects
within these images are comparatively small and dense, following the YOLOv8 strategy of
using the three-level PAFPN, suitable for detecting small objects in images, would limit the
effectiveness of detecting a large number of tiny and densely overlapping objects within
large-sized images. Therefore, we expanded the PAFPN module to a four-layer structure,
taking feature maps from the backbone’s stem layer and stage layers and outputting four
feature maps of the same size to four different detecting heads to detect tiny, small, medium,
and large sizes targets.
As shown in Figure 7, we propose sandwich-fusion (SF), a novel fusion module of
a three-size feature map, which optimizes the target’s spatial and semantic information
for detection heads. The module is applied to the neck’s top-down layers. The inspiration
for this module comes from the BiC model proposed in YOLOv6 3.0 [27]. The input of
sandwich-fusion (SF) is shown in the figure, including feature maps from the backbone’s
lower stage, corresponding stage, and higher stage. The goal is to balance spatial informa-
tion of low-level features and semantic information of high-level features to optimize the
network head’s recognition of the target position and classification.
Drones 2023, 7, 526 9 of 23

Figure 7. Structure of sandwich-fusion (SF).

The sandwich-fusion (SF) module differs from the BiC model of YOLOv6 3.0 in that
it implements a depthwise separable convolution layer [40] to extract the feature map
from the lower stage layer, which contains the target’s most detailed and precise position
information. The module concatenates the lower, corresponding, and upsampled high-level
feature maps as its output. The subsequent C2f module in the top-down layer extracts
this combined information, which is rich in detail regarding the target’s classification and
spatial positioning.
The depthwise separable convolution is a two-stage convolution, which includes
depthwise convolution only for feature maps in a single channel and pointwise convolution
only for feature map channels. Compared to convolutional layers with the same-sized
kernels, the parameters of depthwise separable convolution are relatively small. This
method achieved good results on RepLKNet [41] by expanding the receptive field without
increasing numerous parameters.
The PAFPN module is deployed in the improved YOLOv8 as its neck. Its classification
information was obtained well through the SPPF layer, but spatial information may be
inaccurate. Note that feature maps in the early stages of the backbone are ideal feature
sources for spatial information, as these features do not undergo multiple downsamplings
and retain relatively detailed positional information. However, the disadvantage of these
features is that in the early stage of the backbone, the receptive field of the network is small,
and there are some defects in correctly identifying the background and location information
from complex texture features. Therefore, in the sandwich-fusion (SF) module, a large
kernel depthwise separable convolution is deployed to extract more accurate position
information from the early-stage features and filter out irrelevant information related to
the detected object.

3.3. Proposed Models

In the network neck, the sandwich-fusion (SF) module in the top-down branch of the
extended PAFPN structure varies in the sizes of the convolutional kernels used for the
underlying depthwise separable convolutional layer, depending on the location of different
layers. The specific data are displayed in Table 1:

Table 1. Kernel and channel size of depthwise separable convolutional layers in top-down branch of
proposed Drone-YOLO (large) Model’s neck.

Position of
Module Name Kernel Size of DWConv Channels of DWConv
Module
Top-Down layer 1 Sandwich-fusion 33 32
Top-Down layer 2 Sandwich-fusion 15 64
Top-Down layer 3 Sandwich-fusion 7 128
Drones 2023, 7, 526 10 of 23

Building on the Drone-YOLO (large) model, we employed techniques to create a

range of network models with different scales. These techniques include reducing the C2f
modules in each stage of the backbone, decreasing the channels of convolutional layers.
Consequently, we developed other four distinct scale network models: Drone-YOLO
(medium), Drone-YOLO (small), Drone-YOLO (tiny), and Drone-YOLO (nano).
Among these models, Drone-YOLO (large) achieves the highest detection accuracy but
exhibits the largest size parameters and highest computational complexity. On the other
hand, Drone-YOLO (small) and Drone-YOLO (tiny) possess fewer parameters, deliver
satisfactory performance, and operate at faster processing speeds. Notably, when tested
on UAV remote sensing images, the tiny model performs exceptionally well in detecting
small and tiny targets. It yields results that are comparable to YOLOv8l while utilizing
only one-tenth of the latter’s parameters. This characteristic makes it highly suitable for
deployment on edge computing devices.
For the proposed set of network models, the model parameters for the backbone are
presented in Table 2, while the model parameters for the neck are displayed in Table 3.

Table 2. Proposed model parameters in the backbone.

Stage Layer 1 Stage Layer 2 Stage Layer 3 Stage Layer 4

Model Type Stem SPPF
RepVGG C2f RepVGG C2f RepVGG C2f RepVGG C2f
Drone-YOLO (large) 64 128 128/n = 3 256 256/n = 6 512 512/n = 6 1024 1024/n = 3 1024
Drone-YOLO (medium) 48 96 64/n = 2 192 128/n = 4 384 256/n = 4 768 768/n = 2 768
Drone-YOLO (small) 32 64 64/n = 1 128 128/n = 2 256 256/n = 6 512 512/n = 1 512
Drone-YOLO (tiny) 24 48 48/n = 1 88 88/n = 2 176 176/n = 2 344 344/n = 1 334
Drone-YOLO (nano) 16 32 32/n = 1 64 88/n = 2 128 128/n = 2 256 256/n = 1 256

Table 3. Proposed model parameters in the neck.

Top-Down Branch Bottom-Up Branch

Bottom- Bottom- Bottom-
Model Type Top-Down Layer 1 Top-Down Layer 2 Top-Down Layer 3 Up Layer Up Layer Up
0 1 Layer 2
SF C2f SF C2f SF C2f C2f C2f C2f
Drone-YOLO (large) 32 128/n = 3 64 256/n = 3 128 512/n = 3 256/n = 3 512/n = 3 1024/n = 3
Drone-YOLO (medium) 24 96/n = 2 48 192/n = 2 96 384/n = 2 192/n = 2 384/n = 2 768/n = 2
Drone-YOLO (small) 16 64/n = 1 32 128/n = 1 64 256/n = 1 128/n = 1 256/n = 1 512/n = 1
Drone-YOLO (tiny) 16 48/n = 1 24 88/n = 1 48 176/n = 1 88/n = 1 176/n = 1 344/n = 1
Drone-YOLO (nano) 8 32/n = 1 16 64/n = 1 32 128/n = 1 64/n = 1 128/n = 1 256/n = 1

Table 2 presents the parameters of modules of the backbone, including the stem, stage
layers 1 to 4, and the SPPF. For instance, in the case of Drone-YOLO (large), the stem
consists of a convolutional layer with a channel of 64 and a stride of 2. In stage layer 1,
the downsampling layer is composed of a RepVGG module with a channel of 128 and a
stride of 2. The network component of stage layer 1 comprises 3 cascaded C2f components
with a channel of 128.
Table 3 presents the parameters of the modules within the neck, including top-down
and bottom-up branches. For instance, in the case of the Drone-YOLO (tiny) network, layer 1
of the top-down branch consists of the sandwich-fusion (SF) module’s depthwise separable
convolution with a channel of 16, and a single-level C2f module with a channel of 48.
The Drone-YOLO models utilize the same detection decoupled head as YOLOv8,
where the classification and regression tasks are separate branches. Each branch of the
detection head consists of three cascaded convolutional layers. The first and second layers
employ a 3 × 3 kernel, while the third utilizes a 1 × 1 convolutional kernel. The two
branches are responsible for BBox loss and class loss, respectively. The classification task
Drones 2023, 7, 526 11 of 23

employs binary cross entropy (BCE) loss, while the regression task utilizes a combination
of complete IoU (CIoU) loss and distribution focal loss (DFL).

4. Experiments and Results

4.1. Dataset and Experiments Environment
The VisDrone2019 dataset [42] was collected, annotated, and organized byby the
AISKYEYE team from Tianjin University’s machine learning data mining laboratory. The ob-
ject detection dataset of the VisDrone2019 consists of images and corresponding annotation
files, including 6471 images in the training set, 548 images in the validation set, 1610 images
in the test set, and 1580 images in the competition set. In the dataset, the images sizes varied
from 2000 × 1500 to 480 × 360. Because it is taken from the perspective of the drone, the im-
age is quite different from the image taken by the ground personnel, such as MS-COCO and
VOC2012 in the photo angle, image content, background, and ambient illumination. There
are sample images of VisDrone2019 shown in Figure 8. The scene consisting of dataset
images is extensive, covering streets, squares, parks, schools, residential communities, etc.
The illumination conditions for images include daytime with sufficient lighting conditions,
nighttime with insufficient lighting conditions, cloudy, high-intensity light, and glare con-
ditions. The object types annotated in the image include 10 types: pedestrians, people,
bicycles, cars, vans, trucks, tricycles, awning-tricycle, buses, and motors.
In the experiment, we used Ubuntu 20.04 as the operating system with Python 3.8,
PyTorch 1.16.0, and Cuda 11.6 as the desktop computational software environment. The
experiment utilized NVIDIA 3080ti graphics cards as hardware. The implementation code
of the neural network was modified based on the Ultralytics 8.0.105 version. The hyper-
parameters used during the training, testing, and validation of the experiment remained
consistent. The training epoch was set at 300, and the images inputted into the network
were rescaled to 640 × 640. In the results listed below, all the YOLOv8 and our proposed
Drone-YOLO networks have detection results from our experiments. In these experiments,
none of these networks used pre-training parameters. The remaining data come from
relevant cited papers.
In the embedded application experiment, we used the NVIDIA Tegra TX2 as the
experimental environment, which has a 256-core NVIDIA Pascal architecture GPU, pro-
viding a peak computing performance of 1.33 TFLOPS, and 8GB of memory. The software
environment was Ubuntu 18.04 LTS operating system, NVIDIA JetPack 4.4.1, CUDA 10.2,
and cuDNN 8.0.0.

4.2. Experiment Metrics

The experiments evaluate proposed methods in terms of detection performance and
model parameter size. The experiment metrics include precision (P), recall (R), average
precision (AP), mean average precision (mAP), and million parameters (M) for the network
parameter size.
Precision (P) is the proportion of correctly predicted targets among all detected targets.
It is calculated through Equation (1), where TP represents the correct prediction targets,
and FP represents the incorrect prediction targets.

TP
P= (1)
( TP + FP)

Recall (R) is the proportion of correct detected targets among all existing targets. It is
calculated through Equation (2), where FN represents targets that exist but have not been
correctly detected.
TP
R= (2)
( TP + FN )
Drones 2023, 7, 526 12 of 23

Figure 8. Samples of VisDrone2019: (a) high-altitude aerial view of urban roads at night with low
illumination conditions; (b) residential area images under high- intensity daytime lighting conditions;
(c) urban road intersections under glare conditions; (d) city square scene from a wide view under
cloudy conditions; (e) aerial view image of an urban road area under cloudy conditions; (f) daytime
low-altitude sports stadium scene.

The average precision (AP) represents the area enclosed by the curve formed by the
precision and recall. It is calculated through Equation (3). Our metrics include three
different average precision indicators: AP0.5 , AP0.75 , and AP0.95 . For AP0.5 , in order to
evaluate bounding box prediction as true, the intersection of the union score (IoU) between
the predicted and the annotated bounding box must be higher than 0.50. For AP0.75 ,
bounding box predictions with IoU scores higher than 0.75 are considered to be true.
For AP0.95 , we calculate the average precision values of different IoU scores that vary
within the range of 0.50:0.05:0.95, and then use the average of these calculated average
precision values.
Z 1
AP = p(r )dr (3)
0
The mean average precision (mAP) is the average accuracy of all types of samples.
It is calculated through Equation (4). Our metrics include three different mean average
precision indicators: mAP0.5 , mAP0.75 , and mAP0.95 .

1 k
k i∑
mAP = APi (4)
=1
Drones 2023, 7, 526 13 of 23

4.3. Comparison with the Baseline Methods

Our experimental results were compared with the results of other methods published
on this dataset throughout the years, including faster R-CNN [6], RetinaNet [43], cascade
R-CNN [44], CenterNet [45], improved-YOLOv4 [46], MSA-YOLO [29], TPH-YOLOv5 [47],
and and Li et al. [34]. These methods, as well as YOLOv8-s and YOLOv8-l, are the baseline
methods in this experiment.
The experimental results of the comparison methods shown in Figure 9 were tested
on the VisDrone2019-test dataset by scaling the input image size to 640 × 640. The metric
for the evaluated result is AP0.5 for each object type, and mAP0.5 for all type objects.
The proposed Drone-YOLO (large) achieved the best result in mAP0.5 , and was the best in
seven different type objects. Compared with TPH-YOLOv5 [47], which performed best in
baseline methods, our Drone-YOLO (large) increased the value of mAP0.5 by 9.1%. Drone-
YOLO (medium) and Drone-YOLO (small) with smaller parameter sizes also performed
well compared to baseline methods.

Figure 9. Experimental results for all type object detection on the VisDrone2019-test.

Table 4 shows the performance of the proposed methods and the baseline methods
on the VisDrone2019-test dataset and model parameter size. In our proposed models,
both Drone-YOLO (nano) and Drone-YOLO (tiny) are models with few parameters. How-
ever, Drone-YOLO (tiny) only used 55.4% (regarding the size) of the baseline method of
Li et al.’s [34] parameters and achieved the same result on mAP0.5 .

Table 4. Experimental results on the VisDrone2019-test.

Method mAP0.5 mAP0.75 mAP0.95 Parameters

MobileNetv2-
4.2 - - 3.94 M
SSD
YOLOv4-s 26.5 - - 9.12 M
YOLOv5-s 31.0 - - 9.12 M
YOLOv5-m 33.7 - - 25.1 M
YOLOX-s 26.3 - - 8.94 M
YOLOv7-tiny 26.8 - - 6.02 M
YOLOv8-s [20] 30.9 - - 11.1 M
YOLOv8-l [20] 35.9 21.8 21.2 76.7 M
Li et al. [34] 33.7 - - 9.66 M
Drone-YOLO (nano) 31.0 17.6 17.5 3.05 M
Drone-YOLO (tiny) 33.7 19.4 19.1 5.35 M
Drone-YOLO (small) 35.6 20.6 20.4 10.9 M
Drone-YOLO (medium) 38.9 22.9 22.5 33.9 M
Drone-YOLO (large) 40.7 24.3 23.8 76.2 M
Drones 2023, 7, 526 14 of 23

Table 5 shows the performance of the proposed methods and the baseline methods
on the VisDrone2019-val dataset and model parameter size. Our proposed Drone-YOLO
(large) performs best on mAP0.95 , while MS-YOLOv7 [31] performs best on mAP0.5 .

Table 5. Experimental results on VisDrone2019-val.

Method mAP0.5 mAP0.75 mAP0.95 Parameters

YOLOv7-l [19] 47.1 - 26.4 71.4 M
YOLOv8-l [20] 43.7 27.7 26.9 76.7 M
ACAM-YOLO [30] 49.5 - 29.6 15.9 M
MS-YOLOv7 [31] 53.1 - 31.3 79.7 M
Li et al. [34] 42.2 - - 9.66 M
EdgeYOLO [37] 44.8 26.2 26.4 40.5 M
Drone-YOLO (nano) 38.1 22.8 22.7 3.05 M
Drone-YOLO (tiny) 42.8 26.2 25.6 5.35 M
Drone-YOLO (small) 44.3 27.7 27.0 10.9 M
Drone-YOLO (medium) 48.6 31.1 30.1 33.9 M
Drone-YOLO (large) 51.3 33.2 31.9 76.2 M

In the embedded environment, we evaluated the performance of our proposed meth-

ods, Drone-YOLO (nano), and Drone-YOLO (tiny), as well as the comparative methods,
YOLOv5-n, YOLOv5-s, YOLOv8-n, and YOLOv8-s, on the VisDrone2019-val dataset for
the object detection task. The evaluations were conducted on the NVIDIA Tegra TX2 plat-
form. As shown in Table 6, Drone-YOLO (tiny) achieved the highest detection accuracy on
the dataset.

Table 6. Experimental results on the NVIDIA Tegra TX2, VisDrone2019-val dataset.

Average
Method mAP0.5 mAP0.95 Inference Time Parameters
per Image
YOLOv5-n [16] 33.1 19.2 27.4 ms 2.50 M
YOLOv5-s [16] 39.3 23.4 56.8 ms 9.12 M
YOLOv8-n [20] 34.0 19.8 30.8 ms 3.00 M
YOLOv8-s [20] 39.5 23.5 62.1 ms 11.1 M
Drone-YOLO (nano) 38.1 22.7 127.4 ms 3.05 M
Drone-YOLO (tiny) 42.8 25.6 189.5 ms 5.35 M

4.4. Ablation Experiment

In ablation experiments, the baseline method for the comparison is the YOLOv8-l
model. We added a small-sized object detection head (YOLOv8-l + header), a sandwich
fusion module (YOLOv8-l + sandwich + header), and RepVGG modules in the backbone
as the downsampling layers (YOLOv8-l + sandwich + header + RepVGG). After the above
three stages of improvement, we developed our proposed Drone-YOLO (large) model.
Each model measured multiple metrics in the VisDrone2019-Val dataset. In this experiment,
all hyperparameters remained unchanged. During our training process, the input image
size was 640 × 640, the batch size of all models was set to 8, and the training period was
300 epochs.
The experimental results are shown in Table 7. From the results, the most significant
improvement of the YOLOv8-l model was the addition of small-sized object detection
heads, which improved the most on the strictest mAP0.95 index, reaching 19.3%. This
indicates the importance of detecting small-sized targets from spatial-rich feature maps in
this task. After adding the sandwich fusion module, the improvement measured on the
dataset was the least significant, with a slight decrease in the mAP0.75 indicator. However,
the mAP0.5 and the strictest mAP0.95 indicators slightly increased. When using the RepVGG
Drones 2023, 7, 526 15 of 23

structure as the downsampling module for the backbone part, there was some improvement
in this result, with an increase of 1.2% based on the mAP0.5 indicator.

Table 7. Ablation experiment result in VisDrone2019-val.

Method mAP0.5 mAP0.75 mAP0.95

YOLOv8-l 43.7 27.7 26.4
YOLOv8-l + header 50.5 (↑ 15.6%) 32.7 (↑ 18.1%) 31.5 (↑ 19.3%)
YOLOv8-l + sandwich + header 50.7 (↑ 0.4%) 32.5 (↓ 0.6%) 31.6 (↑ 0.3%)
YOLOv8-l + sandwich + header + RepVGG 51.3 (↑ 1.2%) 33.2 (↑ 2.1%) 31.9 (↑ 0.9%)

4.5. Visualization
The following is an analysis of the differences in the object detection results between
the proposed Drone-YOLO (large), Drone-YOLO (tiny), and YOLOv8-l in images from
different scenes, lighting conditions, shooting positions, and object types.
As shown in Figure 10, in the middle right part of the image, there are five cars driving
on the street, marked with red boxes. The first car just arrived at the foot of the pedestrian
overpass, and two bright lights can be seen shining on the ground. For these five cars,
Drone-YOLO (large) detected two of them, while the other two methods failed to detect
any of them. In the middle-left part of the image, there are two cars driving on the road,
marked with red boxes. Drone-YOLO (large) detected two vehicles with a high degree of
stability. Drone-YOLO (tiny) also detected two vehicles but with slightly lower durability.
YOLOv8-l did not detect either vehicle.
As shown in Figure 11, the upper part of the picture, circled by a large red box, is
a main road with many cars driving on the road. Due to the high altitude and distance
of aerial photography, the size of the car in the image is relatively tiny. YOLOv8-l only
detected three vehicles in this area of the image. Drone-YOLO (large) and Drone-YOLO
(tiny) both detected most targets. In the center of the picture, there are two cars driving on
the auxiliary road. YOLOv8-l did not detect these two cars either; however, both Drone-
YOLO (large) and Drone-YOLO (tiny) detected these two cars. On the left side of the image,
there is a car on a small road behind a building. For the target exposed in the building gap,
neither YOLOv8-l nor Drone-YOLO (tiny) detected it, but Drone-YOLO (large) detected
the car.
As shown in Figure 12, in a low-illumination night environment, there are two motor-
cycles and two cars driving under the overpass in the red box on the left side of the image.
Drone-YOLO (large) detected one car, two motorcycles, and pedestrians on the motorcycles.
YOLOv8-l detected one car and one motorcycle. Drone-YOLO (tiny) only detected one car.
As shown in Figure 13, the central part of the image is a street scene with sufficient
daytime lighting conditions. The drone is about 60 m above the ground, and the street area
circled in the big red box is far from the drone, with a large number of cars driving on the
road. YOLOv8-l can only detect a small number of cars. Both Drone-YOLO (large) and
Drone-YOLO (tiny) can detect many of these cars.
As shown in Figure 14, the image scene is a traffic intersection. The drone is located
relatively low from the ground, and there is a significant glare on the captured road surface.
There is an entrance to a shopping mall on the corner of the street, where crowds enter and
exit. There is a larger red box here. YOLOv8-l detected one person in this area, Drone-YOLO
(large) detected many people, and Drone-YOLO (tiny) detected fewer people, but more
than YOLOv8-l. The traffic post highlighted in the red box in the figure was mistakenly
detected by YOLOv8-l as a tricycle, while Drone-YOLO (large) did not show any false
alarms. Drone-YOLO (tiny) mistakenly detected tricycles and people.
Drones 2023, 7, 526 16 of 23

Figure 10. A daytime urban street scene with insufficient lighting conditions. The image was
captured from an elevated position exceeding 100 meters above the ground. The primary detection
objects include cars, buses, and vans. (a) Results of YOLOv8-l; (b) results of Drone-YOLO (large);
(c) results of Drone-YOLO (tiny).
Drones 2023, 7, 526 17 of 23

Figure 11. A daytime urban street scene with normal lighting conditions. The image was cap-
tured from an elevated position exceeding 100 meters above the ground. The primary detection
objects include cars, buses, and vans. (a) Results of YOLOv8-l; (b) results of Drone-YOLO (large);
(c) results of Drone-YOLO (tiny).
Drones 2023, 7, 526 18 of 23

Figure 12. A nighttime suburban street scene with insufficient lighting conditions. The image
was captured from an elevated position that is approximately 30 meters above the ground at an
oblique angle. The primary detection objects include cars, vans, motorcycles, and pedestrians.
(a) Results of YOLOv8-l; (b) results of Drone-YOLO (large); (c) results of Drone-YOLO (tiny).
Drones 2023, 7, 526 19 of 23

Figure 13. A daytime urban street scene with ample lighting conditions. The image was cap-
tured from an elevated position approximately 60 meters above the ground at an oblique angle.
The primary detection objects include cars, buses, and pedestrians. (a) Results of YOLOv8-l;
(b) results of Drone-YOLO (large); (c) results of Drone-YOLO (tiny).
Drones 2023, 7, 526 20 of 23

Figure 14. A morning street scene with ample lighting conditions. The image was captured from
a slightly inclined position approximately 15 meters above the ground. The primary detection
objects include cars, pedestrians, motorcycles, people, vans, and trucks. (a) Results of YOLOv8-l;
(b) results of Drone-YOLO (large); (c) results of Drone-YOLO (tiny).
Drones 2023, 7, 526 21 of 23

5. Conclusions
In this article, we propose multi-scale drone image object detection algorithms, called
Drone-YOLO, based on the YOLOv8 model. These algorithms are designed to address
the specific challenges associated with drone image object detection. Given the large
number of drone image scenes and the relatively small number of detection objects, we
made improvements to the neck of the YOLOv8 model. We introduced a three-layer
PAFPN structure to enhance the detection of small-sized objects based on 160 × 160 feature
maps. This improvement significantly enhances the algorithm’s ability to detect small-sized
targets. Furthermore, we incorporated the sandwich-fusion module in each layer of the
neck’s up–down branch. This structure allows the fusion of network features with low-level
features that contain rich object spatial information. We achieve this fusion through the
use of depthwise separable evolution, which incurs small parameters and provides a large
receptive field. In the backbone of the network, we utilize the RepVGG module as the
downsampling layer. The RepVGG module enhances the network’s capacity to learn multi-
scale features and yields improved detection results compared to the convolutional layer.
In our experiments, in the ViSDrone2019-test dataset, our proposed Drone-YOLO
(large) outperforms baseline methods in class object detection in mAP0.5 , and performs
the best in 7 out of 10 classes in AP0.5 metrics. In the VisDrone2019-val dataset, our
approach outperforms other methods in terms of the mAP0.95 metrics. Drone-YOLO (tiny),
with only 5.35 M parameters, achieves close results to the baseline method with 9.66 M
parameters in mAP0.5 metrics on the VisDrone2019-test dataset. On the VisDrone2019-val
dataset, it surpasses the baseline method in mAP0.5 metrics. It is worth noting that our
network improvement methods do not incorporate currently popular attention mechanisms,
indicating potential for further enhancements based on task-specific variations.
The proposed methods consist of models with different parameter sizes, where models
with larger parameters are suitable for post-flight object detection on desktop computing
platforms. Models with smaller parameters have undergone testing on the NVIDIA Tegra
TX2 platform and have demonstrated highly accurate results. Considering the computa-
tional capabilities of NVIDIA’s recently released platforms, such as Jetson Xavier NX and
Jetson AGX Xavier, our proposed approach is well-suited for performing high-accuracy
object detection tasks on high-performance embedded platforms integrated with UAVs.

Funding: This research received no external funding.

Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Simonyan, K.; Zisserman, A. Very deep convolutional networks for large-scale image recognition. arXiv 2014, arXiv:1409.1556.
2. He, K.; Zhang, X.; Ren, S.; Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on
Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 27–30 June 2016; pp. 770–778.
3. Huang, G.; Liu, Z.; Van Der Maaten, L.; Weinberger, K.Q. Densely connected convolutional networks. In Proceedings of the IEEE
Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 4700–4708.
4. Girshick, R.; Donahue, J.; Darrell, T.; Malik, J. Rich feature hierarchies for accurate object detection and semantic segmentation.
In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Columbus, OH, USA, 23–28 June 2014;
pp. 580–587.
5. Girshick, R. Fast r-cnn. In Proceedings of the IEEE International Conference on Computer Vision, Santiago, Chile, 7–13 December
2015; pp. 1440–1448.
6. Ren, S.; He, K.; Girshick, R.; Sun, J. Faster r-cnn: Towards real-time object detection with region proposal networks. Adv. Neural
Inf. Process. Syst. 2015, 28.
7. Ronneberger, O.; Fischer, P.; Brox, T. U-net: Convolutional networks for biomedical image segmentation. In Proceedings of the
Medical Image Computing and Computer-Assisted Intervention—MICCAI 2015: 18th International Conference, Munich, Germany, 5–9
October 2015; Proceedings, Part III 18. Springer: Berlin/Heidelberg, Germany, 2015; pp. 234–241.
8. Zhao, H.; Shi, J.; Qi, X.; Wang, X.; Jia, J. Pyramid scene parsing network. In Proceedings of the IEEE Conference on Computer
Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 2881–2890.
Drones 2023, 7, 526 22 of 23

9. He, K.; Gkioxari, G.; Dollár, P.; Girshick, R. Mask r-cnn. In Proceedings of the IEEE International Conference on Computer Vision,
Venice, Italy, 22–29 October 2017; pp. 2961–2969.
10. Everingham, M.; Van Gool, L.; Williams, C.K.I.; Winn, J.; Zisserman, A. The PASCAL Visual Object Classes Challenge 2007
(VOC2007) Results. Available online: http://www.pascal-network.org/challenges/VOC/voc2007/workshop/index.html
(accessed on 30 June 2023).
11. Everingham, M.; Van Gool, L.; Williams, C.K.I.; Winn, J.; Zisserman, A. The PASCAL Visual Object Classes Challenge 2012
(VOC2012) Results. Available online: http://www.pascal-network.org/challenges/VOC/voc2012/workshop/index.html
(accessed on 30 June 2023).
12. Lin, T.; Maire, M.; Belongie, S.J.; Bourdev, L.D.; Girshick, R.B.; Hays, J.; Perona, P.; Ramanan, D.; Doll’a r, P.; Zitnick, C.L. Microsoft
COCO: Common Objects in Context. arXiv 2014, arXiv:1405.0312. Available online: http://xxx.lanl.gov/abs/1405.0312 (accessed
on 30 June 2023).
13. Redmon, J.; Divvala, S.; Girshick, R.; Farhadi, A. You only look once: Unified, real-time object detection. In Proceedings of the
IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 27–30 June 2016; pp. 779–788.
14. Redmon, J.; Farhadi, A. YOLO9000: Better, faster, stronger. In Proceedings of the IEEE Conference on Computer Vision and
Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 7263–7271.
15. Bochkovskiy, A.; Wang, C.Y.; Liao, H.Y.M. Yolov4: Optimal speed and accuracy of object detection. arXiv 2020, arXiv:2004.10934.
16. Jocher, G.; Stoken, A.; Chaurasia, A.; Borovec, J.; Kwon, Y.; Michael, K.; Changyu, L.; Fang, J.; Skalski, P.; Hogan, A.; et al.
Ultralytics/Yolov5: V6.0—YOLOv5n ’Nano’ Models, Roboflow Integration, TensorFlow Export, OpenCV DNN Support, 2021.
Available online: https://zenodo.org/record/5563715 (accessed on 30 June 2023).
17. Ge, Z.; Liu, S.; Wang, F.; Li, Z.; Sun, J. Yolox: Exceeding yolo series in 2021. arXiv 2021, arXiv:2107.08430.
18. Li, C.; Li, L.; Jiang, H.; Weng, K.; Geng, Y.; Li, L.; Ke, Z.; Li, Q.; Cheng, M.; Nie, W.; et al. YOLOv6: A single-stage object detection
framework for industrial applications. arXiv 2022, arXiv:2209.02976.
19. Wang, C.Y.; Bochkovskiy, A.; Liao, H.Y.M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object
detectors. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Vancouver, Canada, 18–22
June 2023; pp. 7464–7475.
20. Jocher, G.; Chaurasia, A.; Qiu, J. YOLO by Ultralytics. 2023. Available online: https://github.com/ultralytics/ultralytics/blob/
main/CITATION.cff (accessed on 30 June 2023).
21. Chu, X.; Zheng, A.; Zhang, X.; Sun, J. Detection in crowded scenes: One proposal, multiple predictions. In Proceedings of the
IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA, 13–19 June 2020; pp. 12214–12223.
22. Wang, C.Y.; Liao, H.Y.M.; Wu, Y.H.; Chen, P.Y.; Hsieh, J.W.; Yeh, I.H. CSPNet: A new backbone that can enhance learning
capability of CNN. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, Seattle,
WA, USA, 14–19 June 2020; pp. 390–391.
23. Liu, S.; Qi, L.; Qin, H.; Shi, J.; Jia, J. Path aggregation network for instance segmentation. In Proceedings of the IEEE Conference
on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 18–23 June 2018; pp. 8759–8768.
24. Elfwing, S.; Uchibe, E.; Doya, K. Sigmoid-weighted linear units for neural network function approximation in reinforcement
learning. Neural Netw. 2018, 107, 3–11. [PubMed]
25. Glorot, X.; Bordes, A.; Bengio, Y. Deep sparse rectifier neural networks. In Proceedings of the Fourteenth International Conference
on Artificial Intelligence and Statistics, JMLR Workshop and Conference Proceedings, Ft. Lauderdale, FL, USA, 11–13 April 2011;
pp. 315–323.
26. Weng, K.; Chu, X.; Xu, X.; Huang, J.; Wei, X. EfficientRep: An Efficient Repvgg-style ConvNets with Hardware-aware Neural
Network Design. arXiv 2023, arXiv:2302.00386.
27. Li, C.; Li, L.; Geng, Y.; Jiang, H.; Cheng, M.; Zhang, B.; Ke, Z.; Xu, X.; Chu, X. YOLOv6 v3.0: A Full-Scale Reloading. arXiv 2023,
arXiv:2301.05586.
28. Wang, C.Y.; Liao, H.Y.M.; Yeh, I.H. Designing Network Design Strategies Through Gradient Path Analysis. arXiv 2022,
arXiv:2211.04800.
29. Mao, G.T.; Deng, T.M.; Yu, N.J. Object detection in UAV images based on multi-scale split attention. J. Aeronaut 2022, 43, 326738.
30. Li, Z.; Wang, Z.; He, Y. Aerial Photography Dense Small Target Detection Algorithm Based on Adaptive Collaborative Attention
Mechanism. J. Aeronaut 2023, 44, 327944.
31. Zhao, L.; Zhu, M. MS-YOLOv7: YOLOv7 Based on Multi-Scale for Object Detection on UAV Aerial Photography. Drones 2023,
7, 188.
32. Liu, Z.; Lin, Y.; Cao, Y.; Hu, H.; Wei, Y.; Zhang, Z.; Lin, S.; Guo, B. Swin transformer: Hierarchical vision transformer using shifted
windows. In Proceedings of the IEEE/CVF International Conference on Computer Vision, Montreal, BC, Canada, 11–17 October
2021; pp. 10012–10022.
33. Woo, S.; Park, J.; Lee, J.Y.; Kweon, I.S. Cbam: Convolutional block attention module. In Proceedings of the European Conference
on Computer Vision (ECCV), Munich, Germany, 8–14 September 2018; pp. 3–19.
34. Li, Y.; Fan, Q.; Huang, H.; Han, Z.; Gu, Q. A Modified YOLOv8 Detection Network for UAV Aerial Image Recognition. Drones
2023, 7, 304. [CrossRef]
35. Tan, M.; Pang, R.; Le, Q.V. Efficientdet: Scalable and efficient object detection. In Proceedings of the IEEE/CVF Conference on
Computer Vision and Pattern Recognition, Seattle, WA, USA, 13–19 June 2020; pp. 10781–10790.
Drones 2023, 7, 526 23 of 23

36. Han, K.; Wang, Y.; Tian, Q.; Guo, J.; Xu, C.; Xu, C. Ghostnet: More features from cheap operations. In Proceedings of the
IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA, 13–19 June 2020; pp. 1580–1589.
37. Liu, S.; Zha, J.; Sun, J.; Li, Z.; Wang, G. EdgeYOLO: An Edge-Real-Time Object Detector. arXiv 2023, arXiv:2302.07483.
38. Ding, X.; Zhang, X.; Ma, N.; Han, J.; Ding, G.; Sun, J. Repvgg: Making vgg-style convnets great again. In Proceedings of the
IEEE/CVF Conference on Computer Vision and Pattern Recognition, Nashville, TN, USA, 20–25 June 2021; pp. 13733–13742.
39. Chetlur, S.; Woolley, C.; Vandermersch, P.; Cohen, J.; Tran, J.; Catanzaro, B.; Shelhamer, E. cudnn: Efficient primitives for deep
learning. arXiv 2014, arXiv:1410.0759.
40. Chollet, F. Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE Conference on Computer
Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 1251–1258.
41. Ding, X.; Zhang, X.; Han, J.; Ding, G. Scaling up your kernels to 31 × 31: Revisiting large kernel design in cnns. In
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, LA, USA, 18–24 June 2022;
pp. 11963–11975.
42. Du, D.; Zhu, P.; Wen, L.; Bian, X.; Lin, H.; Hu, Q.; Peng, T.; Zheng, J.; Wang, X.; Zhang, Y.; et al. VisDrone-DET2019: The vision
meets drone object detection in image challenge results. In Proceedings of the IEEE/CVF International Conference on Computer
Vision Workshops, Seoul, Republic of Korea 27–28 October 2019.
43. Lin, T.Y.; Goyal, P.; Girshick, R.; He, K.; Dollár, P. Focal loss for dense object detection. In Proceedings of the IEEE International
Conference on Computer Vision, Venice, Italy, 22–29 October 2017; pp. 2980–2988.
44. Cai, Z.; Vasconcelos, N. Cascade r-cnn: Delving into high quality object detection. In Proceedings of the IEEE Conference on
Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 18–23 June 2018; pp. 6154–6162.
45. Zhou, X.; Wang, D.; Krähenbühl, P. Objects as points. arXiv 2019, arXiv:1904.07850.
46. Ali, S.; Siddique, A.; Ateş, H.F.; Güntürk, B.K. Improved YOLOv4 for aerial object detection. In Proceedings of the 2021 29th
Signal Processing and Communications Applications Conference (SIU), IEEE, Istanbul, Turkey, 9–11 June 2021; pp. 1–4.
47. Zhu, X.; Lyu, S.; Wang, X.; Zhao, Q. TPH-YOLOv5: Improved YOLOv5 based on transformer prediction head for object detection
on drone-captured scenarios. In Proceedings of the IEEE/CVF International Conference on Computer Vision, Montreal, BC,
Canada, 11–17 October 2021; pp. 2778–2788.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.