Aerial image target detection technology has essential application value in navigation security, traffic control and environmental monitoring. Compared with natural scene images, the background of aerial images is more complex, and there are more small targets, which puts higher requirements on the detection accuracy and real-time performance of the algorithm. To further improve the detection accuracy of lightweight networks for small targets in aerial images, we propose a cross-scale multi-feature fusion target detection method (CMF-YOLOv5s) for aerial images. Based on the original YOLOv5s, a bidirectional cross-scale feature fusion sub-network (BsNet) is constructed, using a newly designed multi-scale fusion module (MFF) and cross-scale feature fusion strategy to enhance the algorithm's ability, that fuses multi-scale feature information and reduces the loss of small target feature information. To improve the problem of the high leakage detection rate of small targets in aerial images, we constructed a multi-scale detection head containing four outputs to improve the network's ability to perceive small targets. To enhance the network's recognition rate of small target samples, we improve the K-means algorithm by introducing a genetic algorithm to optimize the prediction frame size to generate anchor boxes more suitable for aerial images. The experimental results show that on the aerial image small target dataset VisDrone-2019, the proposed method can detect more small targets in aerial images with complex backgrounds. With a detection speed of 116 FPS, compared with the original algorithm, the detection accuracy metrics mAP0.5 and mAP0.5:0.95 for small targets are improved by 5.5% and 3.6%, respectively. Meanwhile, compared with eight advanced lightweight networks such as YOLOv7-Tiny and PP-PicoDet-s, mAP0.5 improves by more than 3.3%, and mAP0.5:0.95 improves by more than 1.9%.
Citation: Yang Pan, Jinhua Yang, Lei Zhu, Lina Yao, Bo Zhang. Aerial images object detection method based on cross-scale multi-feature fusion[J]. Mathematical Biosciences and Engineering, 2023, 20(9): 16148-16168. doi: 10.3934/mbe.2023721
[1] | Sufang Wu, Hua He, Jingjing Huang, Shiyao Jiang, Xiyun Deng, Jun Huang, Yuanbing Chen, Yiqun Jiang . FMR1 is identified as an immune-related novel prognostic biomarker for renal clear cell carcinoma: A bioinformatics analysis of TAZ/YAP. Mathematical Biosciences and Engineering, 2022, 19(9): 9295-9320. doi: 10.3934/mbe.2022432 |
[2] | Miao Zhu, Tao Yan, Shijie Zhu, Fan Weng, Kai Zhu, Chunsheng Wang, Changfa Guo . Identification and verification of FN1, P4HA1 and CREBBP as potential biomarkers in human atrial fibrillation. Mathematical Biosciences and Engineering, 2023, 20(4): 6947-6965. doi: 10.3934/mbe.2023300 |
[3] | Xinwang Yan, Xiaowen Zhao, Qing Yan, Ye Wang, Chunling Zhang . Analysis of the role of METTL5 as a hub gene in lung adenocarcinoma based on a weighted gene co-expression network. Mathematical Biosciences and Engineering, 2021, 18(5): 6608-6619. doi: 10.3934/mbe.2021327 |
[4] | Ming Zhang, Yingying Zhou, Yanli Zhang . High Expression of TLR2 in the serum of patients with tuberculosis and lung cancer, and can promote the progression of lung cancer. Mathematical Biosciences and Engineering, 2020, 17(3): 1959-1972. doi: 10.3934/mbe.2020104 |
[5] | Xiangyue Zhang, Wen Hu, Zixian Lei, Hongjuan Wang, Xiaojing Kang . Identification of key genes and evaluation of immune cell infiltration in vitiligo. Mathematical Biosciences and Engineering, 2021, 18(2): 1051-1062. doi: 10.3934/mbe.2021057 |
[6] | Bo Wei, Rui Wang, Le Wang, Chao Du . Prognostic factor identification by analysis of the gene expression and DNA methylation data in glioma. Mathematical Biosciences and Engineering, 2020, 17(4): 3909-3924. doi: 10.3934/mbe.2020217 |
[7] | Jie Chen, Jinggui Chen, Bo Sun, Jianghong Wu, Chunyan Du . Integrative analysis of immune microenvironment-related CeRNA regulatory axis in gastric cancer. Mathematical Biosciences and Engineering, 2020, 17(4): 3953-3971. doi: 10.3934/mbe.2020219 |
[8] | Fang Niu, Zongwei Liu, Peidong Liu, Hongrui Pan, Jiaxue Bi, Peng Li, Guangze Luo, Yonghui Chen, Xiaoxing Zhang, Xiangchen Dai . Identification of novel genetic biomarkers and treatment targets for arteriosclerosis-related abdominal aortic aneurysm using bioinformatic tools. Mathematical Biosciences and Engineering, 2021, 18(6): 9761-9774. doi: 10.3934/mbe.2021478 |
[9] | Rongxing Qin, Lijuan Huang, Wei Xu, Qingchun Qin, Xiaojun Liang, Xinyu Lai, Xiaoying Huang, Minshan Xie, Li Chen . Identification of disulfidptosis-related genes and analysis of immune infiltration characteristics in ischemic strokes. Mathematical Biosciences and Engineering, 2023, 20(10): 18939-18959. doi: 10.3934/mbe.2023838 |
[10] | Jian-Di Li, Gang Chen, Mei Wu, Yu Huang, Wei Tang . Downregulation of CDC14B in 5218 breast cancer patients: A novel prognosticator for triple-negative breast cancer. Mathematical Biosciences and Engineering, 2020, 17(6): 8152-8181. doi: 10.3934/mbe.2020414 |
Aerial image target detection technology has essential application value in navigation security, traffic control and environmental monitoring. Compared with natural scene images, the background of aerial images is more complex, and there are more small targets, which puts higher requirements on the detection accuracy and real-time performance of the algorithm. To further improve the detection accuracy of lightweight networks for small targets in aerial images, we propose a cross-scale multi-feature fusion target detection method (CMF-YOLOv5s) for aerial images. Based on the original YOLOv5s, a bidirectional cross-scale feature fusion sub-network (BsNet) is constructed, using a newly designed multi-scale fusion module (MFF) and cross-scale feature fusion strategy to enhance the algorithm's ability, that fuses multi-scale feature information and reduces the loss of small target feature information. To improve the problem of the high leakage detection rate of small targets in aerial images, we constructed a multi-scale detection head containing four outputs to improve the network's ability to perceive small targets. To enhance the network's recognition rate of small target samples, we improve the K-means algorithm by introducing a genetic algorithm to optimize the prediction frame size to generate anchor boxes more suitable for aerial images. The experimental results show that on the aerial image small target dataset VisDrone-2019, the proposed method can detect more small targets in aerial images with complex backgrounds. With a detection speed of 116 FPS, compared with the original algorithm, the detection accuracy metrics mAP0.5 and mAP0.5:0.95 for small targets are improved by 5.5% and 3.6%, respectively. Meanwhile, compared with eight advanced lightweight networks such as YOLOv7-Tiny and PP-PicoDet-s, mAP0.5 improves by more than 3.3%, and mAP0.5:0.95 improves by more than 1.9%.
DNA segregation is a complicated process that is critical for cell proliferation and survival [16,30]. Failures during segregation can result in aberrant DNA contents (aneuploidy), a phenomenon which is prevalent in human cancers [24,28]. The fidelity of chromosome segregation during cell division is monitored by control mechanisms called checkpoints, which ensure that particular criteria are met before moving on irreversibly to the next phase [30].
In mitosis, the Spindle Assembly Checkpoint (SAC; [33]) ensures that all chromosomes are properly attached to spindle microtubules via their kinetochores. Even a single unattached or misattached chromosome is sufficient to keep the checkpoint active and engaged [32,31]. (A human mitotic cell has 46 chromosomes and 92 kinetochores.) In budding yeast (Saccharomyces cerevisiae and Drosophila male), an additional control checkpoint exists to place the correct DNA into the right cell during asymmetric cell division. This regulation is known as the Spindle Position Checkpoint (SPOC), and delays mitotic progression until the spindle is correctly aligned along with the polarity axis [2].
Both SAC and SPOC have prominent similarities, although they constitute different mitotic checkpoints. They broadcast a 'wait' signal to the environment, and rely on turnover of the inhibitor and activator at an organelle (kinetochore for SAC and spindle pole body in SPOC). The SAC integrates signaling information about attachment of the individual kinetochores, which broadcast a 'wait'-signal unless a correct attachment is established. Many core SAC components like Mad2 are recruited to unattached kinetochores, and broadcast a nucleoplasm 'wait'-signal and inhibit Cdc20, the APC/C activator. Upon kinetochore-microtubule attachment, these components are rapidly removed from the kinetochores and APC/C:Cdc20 formation (SAC silencing) is turned on (see Fig. 1A). Similarly, central SPOC components Bf1:Bub2 are localized and regulated at the spindle pole bodies (SPBs), which are broadcasted through out the cytosol and inhibit the downstream pathway, Tem1. Signaling from the SPBs is shut down after correct spindle alignment with the polarity axis is achieved (see Fig. 1B).
The checkpoint mechanisms, SAC and SPOC, are hard to observe experimentally in living cells, due to a number of technical challenges. For example, even a low number of components can have various localizations and states upon which the interactions depend. Another issue is that the average diameter of many proteins at the kinetochore is about 40Å, making connections between them invisible to current microscopy techniques. Likewise, the SPOC protein Tem1 activity, whether it is GTP-or GDP-bound, and Bfa1 protein phosphorylation by various kinases (or lack thereof), are not observable in the wet lab. These limitations can be addressed by employing mathematical models and numerical simulations, which can improve our understanding of the mechanics of cell division. However, mathematical methods can be hindered by combinatorial explosion in the amount of intermediate components (complexes) and explicit representations. Also, the different components often interact nonlinearly in time and space; in the presence of various feedback loops, these interactions lead to phenomena that are difficult to predict [10,0,25,11,38]. A combination of experimental work and rigorous mathematical models was central to exclude some hypothesized SAC checkpoint architectures and to elucidate how the SPOC's elaborate system functions (e.g. [4]). All mathematical models extant in the literature studied either human SAC or yeast SPOC activation at a detailed molecular level or in abstract models to distinguish between different pathways [7,34,14,13,15,21,26,19,9,0,18,23,27,29]. However, none of these models addressed SAC or SPOC silencing. Moreover, no rigorous mathematical analysis of properties such as bifurcation, stability, or the existence of feedback loops has been performed for either SAC or SPOC to date. The research groups of Novak and Tyson work intensively on cell cycle-related modeling for yeast. Their recent small model consists of five reactions, five species, and two ODEs based on Michaelis-Menten kinetics with double-negative and two double-positive feedback loops[39]. Also, the smallest chemical reaction system with bistability in the literature contains four reactions, and two ODEs based on mass-action kinetics [41]. It has double positive and a single negative feedback loops. Unfortunately, no regeneration for the reactants is possible in this model.
The purpose of this paper is fourfold. First, a minimal bistable SAC model for activation and silencing was constructed. The model is based on mass-action kinetics and comprises four reactions, double-negative and two double-positive feedback loops, and two ODEs. It is structurally fully distinct to the known smallest biochemical model [41] and structurally comparable to the yeast mitotic model [39,17]. Second, the same model structure was applied to the SPOC, with both SPBs included. Subsequently, a one-parameter bifurcation was computed for these models, in order to demonstrate the realistic biochemical switches. Eventually, numerical simulations were carried out for the system as a system of ordinary differential equations (ODEs), and also as partial differential equations (PDEs; reaction-diffusion systems) with various parameters.
The reaction rules governing the Spindle Assembly Checkpoint (SAC) system are (cf. Fig. 2A):
|
|
|
||
Initial amount | ||||
APC/C | 0.09 |
[37] | ||
MCC | 0.15 |
[15] | ||
Tem1 | 0.06 |
[4] | ||
Bfa1 | 0.04 |
[4] | ||
Bub2 | 0.04 |
[4] | ||
Diffusion constants | ||||
MCC | 1-20 |
This study | ||
APC/C | 1.8 |
[40] | ||
Cdc20 | 19.5 |
[40] | ||
Mad2 | 5 |
[13] | ||
Environment | ||||
Radius of the kinetochore | 0.1 |
0.01 |
[5] | |
Radius of the cell | 10 |
4 |
[21] | |
Rate constants | ||||
kinetochores or SPBs | 0-92 | 0-2 | [16] | |
|
|
This study | ||
|
|
This study | ||
|
|
This study | ||
|
|
This study |
Cdc20:C-Mad2k1,Unattached Kin →MCC | (1) |
MCC+APC/Ck2⇌k−2MCC:APC/C | (2) |
MCC:APC/C k3,Attached Kin,APC/C→Cdc20:C-Mad2+APC/C | (3) |
The Spindle Position Checkpoint mechanism (SPOC) has very similar reaction rules (see below), but differs significantly in the initial concentrations, rate contacts, and signals of exactly two SPBs (cf. Fig. 4A):
Bub2k1,misaligned SPB →Bfa1:Bub2 | (4) |
Bfa1:bub2+Tem1k2⇌k−2Bfa1:Bub2:Tem1 | (5) |
Bfa1:Bub2:Tem1 k3,aligned SPB,Tem1→Bub2+Tem1 | (6) |
By applying the law of mass-action kinetics, the reaction rules (Eqs.(1-3) and Eqs.4-6) can be translated into sets of time-dependent nonlinear ordinary differential equations (ODEs). The translation is done by computing
Adding a diffusion term as a second spatial derivative transforms the system into one of coupled partial differential equations (PDEs), which is known as a reaction-diffusion system and has the following general form:
∂[Ci]∂t=Di∇2[Ci]⏟Diffusion+Rj({[Ci]};P)⏟Reaction, | (7) |
where
∂[Ci]∂t=Dir2∂∂r(r2∂[Ci]∂r)⏟Diffusion+Rj({[Ci]};P)⏟Reaction. | (8) |
The systems of ODEs were implemented in the freely-available software package XPPAUT[8], and integrated using the Rosenbrock method (stiff solver). The bifurcation analyzes and the related numerical integrations were conducted with AUTO [6] via an XPPAUT interface.
For the spatial simulations, the mitotic cell is considered as a 3-sphere with radius
The reaction-diffusion system of PDEs were solved numerically using MATLAB (MathWorks), and integrated using its predefined function called pdepe-solver, which solves systems of parabolic and elliptic PDEs in one space variable
The pdepe-solver converts the PDEs to ODEs using a second-order accurate spatial discretization based on a fixed set of user-specified nodes [36]. This is done using piecewise non-linear Galerkin (regular case) and implicit Petrov-Galerkin (singular case, second-order accurate). The ordinary differential equations resulting from discretization in space are integrated via the multistep solver ode15s which is a variable order solver based on the numerical differentiation formulas (NDFs)using the numerical Gear method [35]. To check that our results were not influenced by the spatial discretization method used in pdepe, we repeated all simulations for 50,100 and 1000 grid cells.
The wiring diagram of the SAC mechanism (Fig. 2A) was translated into a set of reaction equations (see Methods), which were then translated into a set of coupled ordinary differential equations (ODEs) under the assumption of mass action kinetics for all reactions. It is clear biochemically that the total concentration of Cdc20:Mad2 is constant in the system and can be expressed as [Cdc0:CMad2T] = [Cdc20:CMad2] + [MCC] + [MCC:APC/C]. The same is true for total APC/C, thus [APC/CT] = [APC/C] + [MCC:APC/C]. Also, for simplicity, the total amount of MCC was defined as: [MCCT]=[MCC]+[MCC: APC/C]. Under these assumptions, the reduced system can be easily written as the following nonlinear ODEs:
d[MCCT]dt=k1.U([MadT]−[MCCT])−k3.A[APC][MCC:APC/C]) | (9) |
d[APC/C]dt=−k2[MCC][APC/C]+(k−2+A[APC])[MCC:APC/C] | (10) |
The parameter
d[MCCT]dt=k1.U([MadT]−[MCCT])−k3.A([APC/CT]-[MCC:APC/C])[MCC:APC/C]) | (11) |
a.[MCC:APC/C]2+b.[MCC:APC/C]+c=0 | (12) |
where the parameters
First, one-parameter bifurcation analysis was performed for the nonlinear system (Eq. 11 and 16). The aim is to demonstrate the bistable switch states influencing total MCC, while kinetochores are gradually attached. The simulations were conducted using AUTO software (see Methods). The results in Fig. 2B display a typical S-shape, representing the number of attached kinetochores versus the total concentration of the MCC inhibitor. The stable node points for steady states are shown in solid lines and the unstable saddle points are depicted with dashed lines. Stable and unstable steady states meet at saddle-node bifurcation points, which are indicated by solid circles. At the attached kinetochore number (91.98, i.e. nearly all), the SAC checkpoint switched off and APC/C activated rapidly. The total MCC lowers back to zero as the cell enters anaphase. The switch flipping from the SAC-active state to the SAC-inactive state as the number of attached kinetochores raises is indicated by a black dashed line. Slight movements of the bifurcation curve to the left or right are possible, depending on the values of the parameters
Additionally, the dynamics as the change of concentrations over time were simulated and plotted (Fig. 2C) using XPPAUT (see Methods). The concentration of the APC/C component (Fig. 2C pink line) remains very low, as long as no kinetochores are attached. After approximately 25 minutes, the APC/C activity increases quickly to reach its maximum. This result is consistent with the experimental findings reported in the literature [1,27]. The inhibitor complexes MCC:APC/C that sequester APC/C display exactly the opposite behaviors compared to APC/C (Fig. 2C, brown line). MCC concentration behaves similarly to MCC:APC/C with respect to the difference in the initial amount (Fig. 2C, blue line).
Also interesting is the effect of the diffusion coefficient on the SAC system, particularly since the MCC diffusion constant is unknown. To investigate this, a second-derivative diffusion term was added to the original system (Eq.10-Eq.9). The resulting system of coupled PDEs is known as a reaction-diffusion system:
d[MCCT]dt=Dr2∂∂r(r2∂[MCCT]∂r)+k1.U([MadT]−[MCCT])−k3.A[APC][MCC:APC/C]) | (13) |
d[APC/C]dt=Dr2∂∂r(r2∂[APC/C]∂r)−k2[MCC][APC/C]+(k−2+A[APC])[MCC:APC/C] | (14) |
This system was subjected to the initial conditions given in Table 1, with reflective (Neumann) boundary conditions and equivalent geometry details for the cell and kinetochore as specified in Methods. The reaction-diffusion system (Eq.13-Eq.14) was implemented in MATLAB and simulated with various diffusion constant values (Table 1). No qualitative changes were recorded using a wide range of diffusion coefficients for MCC from the literature and higher; all behave similarly as shown in the typical curves in Fig. 3A. To ensure that the assumptions used to reduce the model had no influence of this result, a full system including three PDEs was re-simulated using various diffusion values for MCC. Again, no effects were observed (Fig. 3B). We conclude that diffusion has no major influence on the SAC model, and that the use of ODEs is in principle sufficient. This is certainly not applicable to other model structures, and cannot be generalized, particularly to high-dimensional systems or using low diffusion constants.
Following the same steps as for SAC analysis as in the previous section, the SPOC wiring diagram (Fig. 4A) can be translated into a set of coupled ODEs. Accordingly, the total concentrations of Bub2, Tem1 and Bfa1:Bub2 can be expressed as: [Bub2T] = [Bub2] + [Tem1] + [Bfa1:Bub2:Tem1], [Tem1T] = [Tem1] + [Bfa1:Bub2:Tem1], and [Bfa1:Bub2T] = [Bfa1:Bub2]+[Bfa1:Bub2:Tem1]. The SPOC system is governed by the following equations, under the aforementioned assumptions (see also Methods):
d[Bfa1:Bub2T]dt=k1.X([Bub2T]−[Bfa1:Bub2])−k3.Y([Tem1T]-[Bfa1:Bub2:Tem1])[Bfa1:Bub2:Tem1T] | (15) |
a.[Bfa1:Bub2:Tem1T]2+b.[Bfa1:Bub2:Tem1T]+c=0 | (16) |
where the parameters
Again, AUTO software was used to find the bifurcation curve (see Methods for details). The bifurcation curve (Fig. 4B) is shown for the number of misaligned SPBs versus the total concentration of the Bfa1:Bub2:Tem1 inhibitor. The system switches its bistable state at the value 1.99; subsequently, SPOC is turned off, and eventually Tem1 is rapidly activated.
The numerical simulations of the ODEs is depicted in Fig. 4C, using a stiff solver in XPPAUT. Tem1 (Fig. 2C pink line) is inactive until both SPBs are correctly aligned, which takes place after about 7 minutes. The inhibitor complexes Bfa1:Bub2 and also Bfa1:Bub2:Tem1 have asymmetric behavior compared to Tem1 (Fig. 4C, gray and brown lines).
In eukaryotic cells, the mitotic control prevents DNA missegregation and aneuploidy. The evolutionarily conserved SAC mechanism guarantees that each chromosome has established its attachment to the spindle apparatus before commencing sister-chromatid separation, while the SPOC mechanism assures correct spindle alignment in some asymmetric cell divisions. The complexity of the mitotic control system arises from its fundamental spatial feature. In SAC, a single unattached or incorrectly attached kinetochore (out of 92) has to inhibit all APC/Cs of the cell and solely after last proper attachment the inhibitor has to be switched off rapidly. This behavior likely implies of a feedback loop contribution. The same is applied for SPOC mechanism with the distinguish in the signal that represents two SPBs.
Mathematical modeling can help to improve our molecular-level understanding of the interplay of SAC as well as SPOC components, allowing an understanding of the requirements that the system has to meet. To that end, a mathematical framework containing all signals in SAC (or SPOC) was studied. The network models were constructed based on biochemical reaction rules, and spatial characteristics, such as diffusion coefficients or cell size. Kinetochores and SPBs act as sensory-driven signals in SAC, and SPOC regulations, respectively. The simulation results (as ODEs or PDEs) were able to capture the desired behavior of both SAC and SPOC. Additionally to the simulations, a crucial feedback loop was built around the core components APC/C and MCC (or Tem1 and Bfa1 in SPOC) and their interplay. A one-parameter bifurcation diagram clearly depicts the bistability of the system and the realistic switch from active to inactive control states.
The presented mathematical models can be extended in future work to include various cell cycle checkpoints. Additionally, this approach will serve as a basis for designing experiments and evaluating novel hypotheses related to mitosis and cell cycle. The results provide systems-level details into the DNA segregation control mechanism and demonstrate that the combination of mathematical analysis with experimental data constitutes a powerful tool for investigation of complex biomedical systems.
The author gratefully acknowledges the visiting fund of the Institute for Numerical Simulation (INS) at Bonn University.
[1] |
D. Christine, A. P. S. Chen, H. J. Christanto, Deep learning for highly accurate hand recognition based on YOLOv7 model, Big Data Cogn. Comput., 7 (2023), 53. https://doi.org/10.3390/bdcc7010053 doi: 10.3390/bdcc7010053
![]() |
[2] |
Y. Zhang, J. Chu, L. Leng, J. Miao, Mask-Refined R-CNN: A network for refining object details in instance segmentation, Sensors, 20 (2020), 1010. https://doi.org/10.3390/s20041010 doi: 10.3390/s20041010
![]() |
[3] |
M. Rostami, S. Forouzandeh, K. Berahmand, M. Soltani, Integration of multi-objective PSO based feature selection and node centrality for medical datasets, Genomics, 112 (2020), 3943–3950. https://doi.org/10.1016/j.ygeno.2020.07.027 doi: 10.1016/j.ygeno.2020.07.027
![]() |
[4] | L. A. Varga, B. Kiefer, M. Messmer, A. Zell, SeaDronesSee: A maritime benchmark for detecting humans in open water, in 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), (2022), 3686–3696. https://doi.org/10.1109/WACV51458.2022.00374 |
[5] |
W. Li, J. Qiang, X. Li, P. Guan, Y. Du, UAV image small object detection based on composite backbone network, Mobile Inf. Syst., 2022 (2022), 11. https://doi.org/10.1155/2022/7319529 doi: 10.1155/2022/7319529
![]() |
[6] | Y. Cheng, H. Xu, Y. Liu, Robust small object detection on the water surface through fusion of camera and millimeter wave radar, in 2021 IEEE/CVF International Conference on Computer Vision (ICCV), (2021), 15243–15252. https://doi.org/10.1109/ICCV48922.2021.01498 |
[7] |
J. Ding, N. Xue, G. S. Xia, X. Bai, W. Yang, M. Y. Yang, et al., Object detection in aerial images: A large-scale benchmark and challenges, IEEE Trans. Pattern Anal. Mach. Intell., 44 (2022), 7778–7796. https://doi.org/10.1109/TPAMI.2021.3117983 doi: 10.1109/TPAMI.2021.3117983
![]() |
[8] |
S. Ren, K. He, R. Girshick, J. Sun, Faster R-CNN: Towards real-time object detection with region proposal networks, IEEE Trans. Pattern Anal. Mach. Intell., 36 (2017), 1137–1149. https://doi.org/10.1109/TPAMI.2016.2577031 doi: 10.1109/TPAMI.2016.2577031
![]() |
[9] | J. Redmon, S. Divvala, R. Girshick, A. Farhadi, You only look once: Unified, real time object detection, in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2016), 779–788. https://doi.org/10.1109/CVPR.2016.91 |
[10] |
M. Liu, X. Wang, A. Zhou, X. Fu, Y. Ma, C. Piao, UAV-YOLO: Small object detection on unmanned aerial vehicle perspective, Sensors, 20 (2020), 2238. https://doi.org/10.3390/s20082238 doi: 10.3390/s20082238
![]() |
[11] |
X. Liang, J. Zhang, L. Zhuo, Y. Li, Q. Tian, Small object detection in unmanned aerial vehicle images using feature fusion and scaling-based single shot detector with spatial context analysis, IEEE Trans. Circuits Syst. Video Technol., 30 (2020), 1758–1770. https://doi.org/10.1109/TCSVT.2019.2905881 doi: 10.1109/TCSVT.2019.2905881
![]() |
[12] | X. Liu, J. Huang, T. Yang, Q. Wang, Improved small object detection for UAV acquisition based on CenterNet, Comput. Eng. Appl., 58 (2022), 96–104. |
[13] | Y. Huang, H. Cui, J. Ma, Y. Hao, Research on an aerial object detection algorithm based on improved YOLOv5, in 2022 3rd International Conference on Computer Vision, Image and Deep Learning & International Conference on Computer Engineering and Applications (CVIDL & ICCEA), (2022), 396–400. https://doi.org/10.1109/CVIDLICCEA56201.2022.9825196 |
[14] |
G. Xu, G. Mao, Aerial image object detection of UAV based on multi-level feature fusion, J. Front. Comput. Sci. Technol., 17 (2023), 635–645. https://doi.org/10.3778/j.issn.1673-9418.2205114 doi: 10.3778/j.issn.1673-9418.2205114
![]() |
[15] |
Z. Liu, X. Zhang, C. Liu, H. Wang, C. Sun, B. Li, et al., RelationRS: Relationship representation network for object detection in aerial images, Remote Sens., 14 (2022), 1862. https://doi.org/10.3390/rs14081862 doi: 10.3390/rs14081862
![]() |
[16] |
J. Chu, Z. Guo, L. Leng, Object detection based on multi-layer convolution feature fusion and online hard example mining, IEEE Access, 6 (2018), 19959–19967. https://doi.org/10.1109/ACCESS.2018.2815149 doi: 10.1109/ACCESS.2018.2815149
![]() |
[17] | R. Sheikhpour, K. Berahmand, S. Forouzandeh, Hessian-based semi-supervised feature selection using generalized uncorrelated constraint, Knowledge-Based Syst., 269, (2023), 110521. https://doi.org/10.1016/j.knosys.2023.110521 |
[18] | T. Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, S. Belongie, Feature pyramid networks for object detection, in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2017), 936–944. https://doi.org/10.1109/CVPR.2017.106 |
[19] | S. Liu, L. Qi, H. Qin, J. Shi, J. Jia, Path aggregation network for instance segmentation, in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, (2018), 8759–8768. https://doi.org/10.1109/CVPR.2018.00913 |
[20] | M. Tan, R. Pang, Q. V. Le, Efficientdet: Scalable and efficient object detection, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, (2020), 10781–10790. https://doi.org/10.1109/CVPR42600.2020.01079 |
[21] | G. Jocher, A. Chaurasia, New YOLOv5 Classification Models, 2022. Available from: https://github.com/ultralytics/yolov5/tree/v6.2. |
[22] | S. Liu, D. Huang, Y. Wang, Learning spatial fusion for single-shot object detection, preprint, arXiv: 1911.09516. |
[23] | J. Redmon, A. Farhadi, YOLO9000: Better, faster, stronger, in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2017), 6517–6525. https://doi.org/10.1109/CVPR.2017.690 |
[24] | Z. Tian, C. Shen, H. Chen, T. He, FCOS: Fully convolutional one-stage object detection, in 2019 IEEE/CVF International Conference on Computer Vision (ICCV), (2019), 9626–9635. https://doi.org/10.1109/ICCV.2019.00972 |
[25] | D. Du, P. Zhu, L. Wen, X. Bian, H. Lin, Q. Hu, et al., VisDrone-DET2019: The vision meets drone object detection in image challenge results, in 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW), (2019), 213–226. https://doi.org/10.1109/ICCVW.2019.00030 |
[26] | T. Y. Lin, M. Maire, S. J. Belongie, J. Hays, P. Perona, D. Ramanan, Microsoft COCO: Common objects in context, in 13th European Conference on Computer Vision, (2014), 740–755. https://doi.org/10.1007/978-3-319-10602-1_48 |
[27] |
Z. Zhang, H. Yi, J. Zheng, Focusing on small objects detector in aerial images, Acta Electron. Sin., 51 (2023), 944–955. https://doi.org/10.12263/DZXB.20220313 doi: 10.12263/DZXB.20220313
![]() |
[28] | M. Tan, Q. Le, Efficientnet: Rethinking model scaling for convolutional neural networks, in International Conference on Machine Learning, PMLR, (2019), 6105–6114. |
[29] | A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, MobileNets: efficient convolutional neural networks for mobile vision applications, preprint, arXiv: 1704.04861. |
[30] | J. Redmon, A. Farhadi, YOLOv3: An incremental improvement, preprint, arXiv: 1804.02767. |
[31] | Z. Ge, S. Liu, F. Wang, Z. Li, J. Sun, YOLOX: Exceeding YOLO series in 2021, preprint, arXiv: 2107.08430. |
[32] | G. Yu, Q. Chang, W. Lv, C. Xu, C. Cui, W. Ji, et al., PP-PicoDet: A better real-time object detector on mobile devices, preprint, arXiv: 2111.00902. |
[33] | C. Y. Wang, A. Bochkovskiy, H. Y. M. Liao, 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, (2023), 7464–7475. |
[34] | S. Xu, X. Wang, W. Lv, Q. Chang, C. Cui, K. Deng, et al., PP-YOLOE: An evolved version of YOLO, preprint, arXiv: 2203.16250. |
1. | Stephan Peter, Fanar Ghanim, Peter Dittrich, Bashar Ibrahim, Organizations in reaction-diffusion systems: Effects of diffusion and boundary conditions, 2020, 43, 1476945X, 100855, 10.1016/j.ecocom.2020.100855 | |
2. | Richard Henze, Chunyan Mu, Mate Puljiz, Nishanthan Kamaleson, Jan Huwald, John Haslegrave, Pietro Speroni di Fenizio, David Parker, Christopher Good, Jonathan E. Rowe, Bashar Ibrahim, Peter Dittrich, Multi-scale stochastic organization-oriented coarse-graining exemplified on the human mitotic checkpoint, 2019, 9, 2045-2322, 10.1038/s41598-019-40648-w | |
3. | Stephan Peter, Bashar Ibrahim, Peter Dittrich, Linking Network Structure and Dynamics to Describe the Set of Persistent Species in Reaction Diffusion Systems, 2021, 20, 1536-0040, 2037, 10.1137/21M1396708 | |
4. | Faiza Hanif Waghu, Karishma Desai, Sumana Srinivasan, Kaushiki S. Prabhudesai, Vikas Dighe, Kareenhalli V. Venkatesh, Susan Idicula-Thomas, FSHR antagonists can trigger a PCOS-like state, 2022, 68, 1939-6368, 129, 10.1080/19396368.2021.2010837 |
|
|
|
||
Initial amount | ||||
APC/C | 0.09 |
[37] | ||
MCC | 0.15 |
[15] | ||
Tem1 | 0.06 |
[4] | ||
Bfa1 | 0.04 |
[4] | ||
Bub2 | 0.04 |
[4] | ||
Diffusion constants | ||||
MCC | 1-20 |
This study | ||
APC/C | 1.8 |
[40] | ||
Cdc20 | 19.5 |
[40] | ||
Mad2 | 5 |
[13] | ||
Environment | ||||
Radius of the kinetochore | 0.1 |
0.01 |
[5] | |
Radius of the cell | 10 |
4 |
[21] | |
Rate constants | ||||
kinetochores or SPBs | 0-92 | 0-2 | [16] | |
|
|
This study | ||
|
|
This study | ||
|
|
This study | ||
|
|
This study |
|
|
|
||
Initial amount | ||||
APC/C | 0.09 |
[37] | ||
MCC | 0.15 |
[15] | ||
Tem1 | 0.06 |
[4] | ||
Bfa1 | 0.04 |
[4] | ||
Bub2 | 0.04 |
[4] | ||
Diffusion constants | ||||
MCC | 1-20 |
This study | ||
APC/C | 1.8 |
[40] | ||
Cdc20 | 19.5 |
[40] | ||
Mad2 | 5 |
[13] | ||
Environment | ||||
Radius of the kinetochore | 0.1 |
0.01 |
[5] | |
Radius of the cell | 10 |
4 |
[21] | |
Rate constants | ||||
kinetochores or SPBs | 0-92 | 0-2 | [16] | |
|
|
This study | ||
|
|
This study | ||
|
|
This study | ||
|
|
This study |