Transfer Network Research Articles

Application of convolutional neural networks in image classification and applications of improved convolutional neural networks

Abstract. This paper reviews the application and improvement of convolutional neural networks (CNNs) in image classification. Firstly, a shallow CNN for interstitial lung disease image classification is presented. This model suppresses overfitting through a unique network architecture and optimisation algorithm. Next, the improved VGG16 architecture and MIDNet18 model are discussed and their superior performance in brain tumour image classification is demonstrated. Subsequently, a CNN-CapsNet model for cervical cancer image classification and its improvement are presented and the customised model is compared with the conventional VGG-16 CNN architecture in the paper. Next, the application of sparse convolutional kernels and hybrid sparse convolutional kernels (HDCs) in solving the problem of computational resource consumption is presented. Subsequently, methods for solving the problem of limited training data through transfer learning and network data augmentation techniques are discussed, as well as GAN-generated datasets for solving the overfitting problem. Finally, the effect of degraded images on the classification effectiveness of CNNs is explored. The results show that the improved CNN architecture and algorithms have significant effects in solving the problems of overfitting and computational resource consumption, and can significantly improve the accuracy and efficiency of image classification. And degraded images do adversely affect the accuracy of CNN for image classification.

The impact of multiplex relationships on households’ informal farmland transfer in rural China: A network perspective

Over the last three decades, farmland transfers in rural China have significantly increased. Despite this rise, informal farmland transfers within social networks continue to be common in many villages, underscoring the need for a more nuanced understanding of their determinants. Existing literature extensively explores the influence of social networks on farmland transfers, it often overlooks the distinction between formal and informal transactions and primarily focuses on binary relationships. Limited data availability and constraints in econometric methodology have impeded quantitative estimation of the impact of farmers’ interactions on informal farmland transfers at the local levels. To address this issue, we developed a framework that integrates embeddedness theory and social interaction theory to uncover the mechanism between multiplex relationship and informal farmland transfers from a network perspective. Using survey data collected from all 341 households in a village in central China, we constructed both relationship networks and informal farmland transfer network among peasant households. We then employed the logistic regression quadratic assignment procedure (LRQAP) to quantify the impact of multiplex relationships and network structures on informal farmland transfers. The results reveal a significant positive correlation between multiplex relationships and informal farmland transfer behaviors among peasant households. Specifically, households are more likely to engage in informal farmland transfer through relationships characterized by emotional connections and reciprocal exchanges, rather than through relationships based on cooperative support. Furthermore, our findings indicate that multiplex relationships influence informal farmland transfer behaviors through network structure effects, such as reciprocity and preferential attachment, rather than through transitivity. These results highlight the importance of leveraging multiplex relationships to facilitate and regulate informal farmland transfers amid rural decline and land use change.

Intelligent Predictive Networks for Cattaneo-Christov Heat and Mass Transfer Dissipated Williamson Fluid Through Double Stratification

This study aims to develop an efficient predictive model for Cattaneo-Christov heat and mass transformation of dissipative Williamson fluid with the effects of double stratification (CCHMT-DWF-DS) using the Levenberg-Marquardt Backpropagation (LMA-BP) algorithm. The under-consideration Williamson fluid flow is magneto-hydro-dynamic, incompressible and two-dimensional through a stretching sheet. The mathematical model of nonlinear partial differential equations for physical phenomena is transformed into ordinary differential equations by means of renowned similarity transformations. The solutions of physical problem are computed by bvp4c technique through MATLAB. The LMA-BP is employed to train a backward neural network capable of accurately predicting velocity, temperature, and concentration profiles under various physical conditions such as changes in the Hartmann number , Prandtl number , Schmidt number , Williamson parameter , the relaxation time of temperature , the relaxation time of concentration , temperature stratification , and concentration stratification for generating a variety of graphical outcomes and statistics. This research is significant for its innovative use of the LMA-BP in analyzing the complex dynamics of non-Newtonian fluids specifically the Williamson fluid, alongside the Cattaneo-Christov heat and mass flux model. The obtaining graphs have been discussed in detail. The thermal and solutal relaxation factors reduce heat and mass flow while fluid motion is delayed by the time-dependent parameter and further reduced by the Hartman number. The Cattaneo-Christov heat flux model enhances simulation accuracy by integrating temporal delays in heat transfer, proving beneficial for sophisticated industrial and scientific endeavors related to non-Newtonian fluids. This analysis offers a powerful predictive tool for applications in thermal management, industrial cooling systems, and biomedical fluid dynamics, advancing machine learning in fluid mechanics.

Few-Shot Face Sketch-to-Photo Synthesis via Global-Local Asymmetric Image-to-Image Translation

Face sketch-to-photo synthesis is widely used in law enforcement and digital entertainment, which can be achieved by Image-to-Image (I2I) translation. Traditional I2I translation algorithms usually regard the bidirectional translation of two image domains as two symmetric processes, so the two translation networks adopt the same structure. However, due to the scarcity of face sketches and the abundance of face photos, the sketch-to-photo and photo-to-sketch processes are asymmetric. Considering this issue, we propose a few-shot face sketch-to-photo synthesis model based on asymmetric I2I translation, where the sketch-to-photo process uses a feature-embedded generating network, while the photo-to-sketch process uses a style transfer network. On this basis, a three-stage asymmetric training strategy with style transfer as the trigger is proposed to optimize the proposed model by utilizing the advantage that the style transfer network only needs few-shot face sketches for training. Additionally, we discover that stylistic differences between the global and local sketch faces lead to inconsistencies between the global and local sketch-to-photo processes. Thus, a dual branch of the global face and local face is adopted in the sketch-to-photo synthesis model to learn the specific transformation processes for global structure and local details. Finally, the high-quality synthetic face photo can be generated through the global-local face fusion sub-network. Extensive experimental results demonstrate that the proposed Global-Local Asymmetric (GLAS) I2I translation algorithm compared to SOTA methods, at least improves FSIM by 0.0126, and reduces LPIPS (alex), LPIPS (squeeze), and LPIPS (vgg) by 0.0610, 0.0883, and 0.0719, respectively.

Multi-scale self-attention feature decoupling transfer network-based cross-domain capacity prediction of lithium-ion batteries

Lithium-ion battery capacity prediction is paramount for improving the safety and reliability of energy storage devices. However, accurately predicting capacity continues to pose a challenge, stemming from the diverse range of operational conditions and the lack of capacity labels under unpredictable scenarios. To tackle this problem, a multi-scale self-attention feature decoupling transfer network is proposed for predicting the capacity of lithium-ion batteries across diverse operational conditions. More specifically, multi-scale multi-head self-attentive encoders are designed to adaptively extract multi-scale domain-invariant features by using a series of loss function constraints. Then, transfer learning integrated with domain-invariant features is applied to migrate the knowledge of the source domain to the target domain for cross-domain capacity prediction. Finally, the performance of our methods is authenticated through utilization of two battery dataset, each including three different charging and discharging scenarios. Experimental results indicate that the proposed method effectively extracts domain-invariant features from charging data characterized by varying distributions, thereby mitigating the effects of distributional differences. Compared with the state-of-the-art methods, the proposed method shows excellent accuracy and bustling ability in two datasets.

Metal-organic framework-derived in situ-grown Cu2Sb@NC octahedra for high-performance stable lithium/sodium storage

Intermetallic compounds are considered a special class of alloy-type anode materials that show great potential for use in lithium/sodium-ion batteries due to their inherent advantages, such as high specific capacity and enhanced safety features. However, their practical application is often impeded by the substantial volume changes that occur during the insertion and extraction of Li+/Na+ ions, leading to capacity decay and reduced service life. To address this challenge, we developed a novel composite material comprising Cu2Sb nanoparticles encapsulated within a N-doped octahedral carbon matrix (Cu2Sb@N-C). This composite structure features a heteroatom-doped conductive network, a three-dimensional carbon framework, and finely dispersed Cu2Sb nanoparticles. Benefiting from the heteroatom-doped conductive network, three-dimensional carbon framework and ultrafine Cu2Sb nanoparticles, they not only provide a penetrating network for the fast transfer of charge carriers and improve the accessibility of ions, but also alleviate the agglomeration and volume expansion of Cu2Sb during cycling. For LIBs, the material exhibited a reversible capacity of 1001.0 mAh/g after 330 cycles at 0.2 A/g, while maintaining a substantial capacity of 565.5 mAh/g after 1000 cycles at 1.0 A/g. In the case of SIBs, the material exhibited a capacity of 375.6 mAh/g after 100 cycles at 0.2 A/g. Post-cycling electrode analyses revealed that the Cu2Sb@N-C anode material retained its structural integrity, demonstrating its ability to withstand the continuous insertion and extraction of Li+/Na+ ions. This research contributes insights into the effective design of innovative high-capacity alloy-based anode materials for advanced secondary batteries.

SCC/PVA Hydrogel with Self‐Healing Capability for Electrochromic Device

AbstractCracking caused by aging or repeated bending can severely affect the lifetime of electrochromic devices. In this work, a self‐healing electrochromic hydrogel was developed with poly (vinyl alcohol) (PVA)‐sodium carboxymethyl chitosan (SCC) semi‐inter transfer network as the matrix, viologen bromide (HPV2+2Br−) as the electrochromic substance, and borax as the cross‐linking agent. The self‐healing properties of hydrogel stem from the dynamic and reversible borate bonds between borax and PVA molecular chains. As the voltage changes, the color of the device transitions from colorless to light purple and then to dark purple. The optical contrast at 600 nm reaches approximately 51 %. At room temperature, the disconnected hydrogel can fully restore its original state after 30 minutes of contact, effectively addressing issues related to cracks and damage in flexible electrochromic devices. This research holds significant implications for enhancing the reliability and stability of flexible electrochromic devices.

Identifying successful football teams in the European player transfer network

This paper considers the European transfer market for professional football players as a network to study the relation between a team’s position in this network and performance in its domestic league. Our analysis is centered on eight top European leagues. The market in each season is represented as a weighted directed network capturing the transfers of players to or from the teams in these leagues, and we also consider the cumulative network over the past 28 years. We find that the overall structure of this transfer market network has properties commonly observed in real-world networks, such as a skewed degree distribution, high clustering, and small-world characteristics. To assess football teams we first construct a measure of within-league performance that is comparable across leagues. Regression analysis is used to relate league performance with both the network position and level of engagement of the team in the transfer market, under two complimentary setups. Network position variables include, e.g., betweenness centrality, closeness centrality and node clustering coefficient, whereas market engagement variables capture a team’s activity in the transfer market, e.g., total number of player transfers and total paid for players. For the season snapshots, the number of transfers correspond to weighted in- and out-degree. Our analysis first corroborates several recent findings relating aspects of market engagement with teams’ league performance. A higher number of incoming transfers indicates worse performance and better resourced teams perform better. Then, and across specifications, we find that network position variables remain salient even when engagement variables are already considered. This substantiates the notion in the existing literature that a high degree corresponds to better team performance and suggests that network aspects of trading strategy may affect a team’s success in their respective domestic league (or vice versa). In this sense, the approach and findings presented in this paper may in the future guide team’s player acquisition policies.

Multi-source domain transfer network based on subdomain adaptation and minimum class confusion for EEG emotion recognition

Electroencephalogram (EEG) signals, which objectively reflect the state of the brain, are widely favored in emotion recognition research. However, the presence of cross-session and cross-subject variation in EEG signals has hindered the practical implementation of EEG-based emotion recognition technologies. In this article, we propose a multi-source domain transfer method based on subdomain adaptation and minimum class confusion (MS-SAMCC) in response to the addressed issue. First, we introduce the mix-up data augmentation technique to generate augmented samples. Next, we propose a minimum class confusion subdomain adaptation method (MCCSA) as a sub-module of the multi-source domain adaptation module. This approach enables global alignment between each source domain and the target domain, while also achieving alignment among individual subdomains within them. Additionally, we employ minimum class confusion (MCC) as a regularizer for this sub-module. We performed experiments on SEED, SEED IV, and FACED datasets. In the cross-subject experiments, our method achieved mean classification accuracies of 87.14% on SEED, 63.24% on SEED IV, and 42.07% on FACED. In the cross-session experiments, our approach obtained average classification accuracies of 94.20% on SEED and 71.66% on SEED IV. These results demonstrate that the MS-SAMCC approach proposed in this study can effectively address EEG-based emotion recognition tasks.

Transferable, highly crystalline covellite membrane for multifunctional thermoelectric systems

AbstractEmerging freestanding membrane technologies, especially using inorganic thermoelectric materials, demonstrate the potential for advanced thermoelectric platforms. However, using rare and toxic elements during material processing must be circumvented. Herein, we present a scalable method for synthesizing highly crystalline CuS membranes for thermoelectric applications. By sulfurizing crystalline Cu, we produce a highly percolated and easily transferable network of submicron CuS rods. The CuS membrane effectively separates thermal and electrical properties to achieve a power factor of 0.50 mW m−1 K−2 and thermal conductivity of 0.37 W m−1 K−1 at 650 K (estimated value). This yields a record‐high dimensionless figure‐of‐merit of 0.91 at 650 K (estimated value) for covellite. Moreover, integrating 12 CuS devices into a module resulted in a power generation of ~4 μW at ΔT of 40 K despite using a straightforward configuration with only p‐type CuS. Furthermore, based on the temperature‐dependent electrical characteristics of CuS, we develop a wearable temperature sensor with antibacterial properties.image

A multi-source domain feature-decision dual fusion adversarial transfer network for cross-domain anti-noise mechanical fault diagnosis in sustainable city

Rotating machinery forms the critical backbone of infrastructure in a sustainable city, with bearings playing a pivotal role as key mechanical transmission components. Therefore, the health status of these bearings directly influences the safe operation of the infrastructure. Accurate and reliable diagnosis of defects in these components minimizes downtime, reduces maintenance costs, and prevents major accidents, ultimately providing insights in the construction and management of a sustainable city. Typically, in actual industrial scenarios, varying working conditions and various types of machines can result in significant discrepancies in the distribution of sample data. Moreover, the non-negligible noise may degrade the diagnostic performance. Therefore, realizing an accurate and reliable bearing diagnosis considering the cross-domain and noise environment remains a challenge. Leveraging the merits of information fusion and multi-source domain transfer learning, this article proposes a multi-source domain feature-decision dual fusion adversarial transfer network (DFATN) to break through the aforesaid limitations. Initially, an adversarial transfer framework is developed, incorporating novel feature matching evaluation and joint distribution difference losses. This framework is designed to facilitate the learning of feature invariants across domains and to enhance the sharing of domain-specific knowledge, even in noise. Relying on channel-spatial interactive feature fusion, a multi-scale feature extractor (MFE) is constructed to share the interaction and enhance the modeling of complex features in multiple dimensions. Additionally, a fault state-related decision fusion mechanism (SDF) is also implemented to integrate diagnostic information, significantly enhancing the generalization performance and robustness of the proposed network. By employing both public Paderborn University (PU) and laboratory-collected (Lab) datasets, the effectiveness and superiority of the proposed DFATN on bearing fault diagnosis are validated. For cross-working condition tasks, the proposed method realizes impressive performance, with average accuracies of 96.52% and 98.76% for Paderborn University (PU) and laboratory-collected (Lab) datasets, respectively. For cross-machine tasks, the average accuracy is 83.36%, outperforming other latest cross-domain fault diagnosis techniques.

Design principles for energy transfer in the photosystem II supercomplex from kinetic transition networks

Photosystem II (PSII) has the unique ability to perform water-splitting. With light-harvesting complexes, it forms the PSII supercomplex (PSII-SC) which is a functional unit that can perform efficient energy conversion, as well as photoprotection, allowing photosynthetic organisms to adapt to the naturally fluctuating sunlight intensity. Achieving these functions requires a collaborative energy transfer network between all subunits of the PSII-SC. In this work, we perform kinetic analyses and characterise the energy landscape of the PSII-SC with a structure-based energy transfer model. With first passage time analyses and kinetic Monte Carlo simulations, we are able to map out the overall energy transfer network. We also investigate how energy transfer pathways are affected when individual protein complexes are removed from the network, revealing the functional roles of the subunits of the PSII-SC. In addition, we provide a quantitative description of the flat energy landscape of the PSII-SC. We show that it is a unique landscape that produces multiple kinetically relevant pathways, corresponding to a high pathway entropy. These design principles are crucial for balancing efficient energy conversion and photoprotection.

MGACL: Prediction Drug-Protein Interaction Based on Meta-Graph Association-Aware Contrastive Learning.

The identification of drug-target interaction (DTI) is crucial for drug discovery. However, how to reduce the graph neural network's false positives due to its bias and negative transfer in the original bipartite graph remains to be clarified. Considering that the impact of heterogeneous auxiliary information on DTI varies depending on the drug and target, we established an adaptive enhanced personalized meta-knowledge transfer network named Meta Graph Association-Aware Contrastive Learning (MGACL), which can transfer personalized heterogeneous auxiliary information from different nodes and reduce data bias. Meanwhile, we propose a novel DTI association-aware contrastive learning strategy that aligns high-frequency drug representations with learned auxiliary graph representations to prevent negative transfer. Our study improves the DTI prediction performance by about 3%, evaluated by analyzing the area under the curve (AUC) and area under the precision-recall curve (AUPRC) compared with existing methods, which is more conducive to accurately identifying drug targets for the development of new drugs.

Characterization of Industrial Carbon Emission Dynamics and Network Structure Evolution in China

Given that industry is the major source of carbon emissions, clarifying the carbon transfer regularity triggered by industry production linkage is of great importance in realizing the synergistic development of industry to reduce pollution and carbon emissions. Based on the perspective of responsibility allocation, the study accounted for the industry carbon emissions and constructed a carbon transfer weighted directed network model and applied the social network analysis method to explore the structure of the industry carbon transfer network and its evolution characteristics from 2005 to 2020. The results showed that： ① The overall carbon emissions of industries from 2005 to 2020 presented a rapid growth trend and large differences occurred in emissions between industries. ② The scale of the carbon transfer network was expanding rapidly and the characteristics of the small-world network were significant. ③ The community structure of the carbon transfer network was obvious, including four optimal associations： high carbon emission, carbon transmission, carbon endogenous, and high spillover. ④ The construction industry and the metal smelting and rolling processing industry were the core of the carbon transfer network structure, with strong network control. In view of the above characteristics, we proposed countermeasures and suggestions in terms of implementing the allocation of carbon emission reduction responsibilities and differentiated emission reduction measures for associations as well as focusing on the management of core nodes of the network, with the intent of providing cross-sectoral synergistic emission reduction ideas for the development of sustainable low-carbon transformation in China's industry sector.

Research on Intelligent Digital Manufacturing Platform Based on Artificial Intelligence Deep Neural Network

In today’s information age, intelligent manufacturing has become a significant trend of future industrial development. The neural network plays a vital role as the basis of smart manufacturing. This is a kind of intelligent algorithm that can simulate human brain organization, with strong learning ability and computing power, and has achieved remarkable results in the application field; the most typical representative is deep neural network, especially in recent years, an artificial intelligence developed based on deep neural network has entered the application stage, providing a new idea and model for industrial production. Based on this, taking the clothing production process as an example, this paper combined transfer learning and network fine-tuning technology to build a Faster RCNN model based on the ResNet-101 network to realize the positioning and recognition of clothing style map features and create an intelligent generation platform for clothing process flow on this basis.

Solution-processed, ultrasensitive, high current density vertical phototransistor using porous carbon nanotube electrode

Multi-wall carbon nanotube (MWCNT) thin film has been identified as the ideal electrode material due to their inherent electronic and optoelectronic properties. We have successfully fabricated a vertical field-effect phototransistor (vFEpT) using MWCNT as a porous source electrode. The optimized MWCNT electrodes showed an electrical conductivity of 3.92 × 103 S/m and a sheet resistance of about 490 Ω/□, with a layer thickness of about 520 nm. The as-fabricated transistor has p-type conductivity because of the unintentional doping of the PbS layer and the intrinsic hole transport property of the MWCNT network. Additionally, the device shows excellent photoelectric performance at 30 mWcm−2 of illumination, with a high current density of 2.3 × 10−3 Acm−2. With a high conductivity network for carrier transfer and separation, the special characteristics of MWCNT nanoporous thin film enable improved photoelectric properties, such as a high photoresponsivity of 3.16 AW−1 and a specific detectivity of 2.33 × 1012 Jones when illuminated at 808 nm light irradiance with an intensity of 0.6 µW/cm2. Furthermore, the device demonstrated a steady dynamic response to an optical signal, with rising and falling times of 31.8 ms and 32.9 ms, respectively. This work presents a new method for creating high-performance vertical phototransistors with MWCNT as the source electrode.

Power distribution system restoration based on soft open points and islanding by distributed generations

Abstract A powerful and efficient program for restoring the electrical distribution system by effectively utilizing the maximum capabilities within the system, including soft open points, distributed generation resources, and network configuration changes, can significantly reduce both the quantity and duration of lost loads caused by permanent faults. In this research study, we address the issue of distribution network restoration with a focus on soft open points (SOPs) and intentional islanding using distributed generations. This problem is formulated as a constrained optimization problem in order to determine the optimal distribution network configuration, quality of intentional islanding through DGs, control function of SOPs, and amount of load shedding. The objective is to minimize lost load while minimizing switching operations and preferably minimal islanding. Given that this problem involves multiple complexities such as combinatorial nature, mixed-integer variables, non-linearity, non-convexity along with numerous variables and constraints; we employ an evolutionary method known as simulated annealing (SA) algorithm to solve it without any simplifications or assumptions about convexity. To enhance efficiency during implementation of SA algorithm, Kruskal’s algorithm is utilized for generating radial solutions which restricts search space to feasible solutions resulting in quicker attainment of high-quality optimal solutions. Finally, the outcomes of implementing the suggested approach on the 69-bus IEEE distribution system are presented and examined. It is demonstrated that by leveraging the potential of soft open points in load transfer control and network voltage regulation, along with utilizing intentional islanding capability provided by distributed generation resources, the restoration capability of the distribution system can be greatly enhanced.

Research on parameter deviation modelling of dual LCC wireless power transfer system

Abstract In this paper, parameter deviation modelling analysis is performed based on the dual LCC compensation network of wireless power transfer (WPT) system. Firstly, the mutual inductive coupling model of the dual LCC compensation circuit is derived based on the fundamental wave analysis method; secondly, the parameter deviation is introduced into the circuit model for the errors existing in the inductive and capacitive components of the compensation network circuit of the actual wireless charging system, and the relation between the deviation and the amplitude of the branch currents, the phase angle and the output characteristics of system are obtained through the establishment of derivation model and the resistive deviation correction term is introduced; finally, the parameter deviation is verified through the simulation. The establishment of parameter deviation model increases the accuracy of the whole circuit model and provides the theoretical basis to reduce the negative influence of parameter deviation on transmission characteristics. The value of parameter deviation can be inverted through the phase angle, thus providing a basis for adjusting the resonant state of the compensation circuit.

ANI/EFP: Modeling Long-Range Interactions in ANI Neural Network with Effective Fragment Potentials.

Deep learning Neural Networks (NN) have been developed in the field of molecular modeling for the purpose of circumventing the high computational cost of quantum-mechanical calculations while rivaling their accuracies. Although these networks have found great success, they generally lack the ability to accurately describe long-range interactions, which makes them unusable for extended molecular systems. Herein, we provide a method for partially retraining the deep learning general-use neural network ANI, in which the long-range interactions are represented via atomic electrostatic potentials. The electrostatic potentials, generated with polarizable effective fragment potentials (EFP), are used as an additional input feature for the network. This new ANI/EFP network can predict solute-solvent interaction energies on a trained data set with a kcal/mol accuracy. It also shows promise in predicting the interaction energies of a solute in solvent environments that have not been included in a training data set. The proposed protocol can be taken as an example and further developed, leading to highly accurate and transferable neural network potentials capable of handling long-range interactions and extended molecular systems.

Deep Learning Based Method for Dynamic Tracking of Bubbles in Microfluidic Gas-Driven Water Experiments

Using microfluidic chips for gas displacement experiments is an important approach for studying CO2 displacement mechanisms and gas-liquid exchange processes. Analyzing the micro-scale distribution of gas and water during gas displacement enables the optimization of injection strategies, offering valuable guidance for enhancing oil and gas recovery rates. Traditional image processing techniques often face challenges in effectively handle the complex gas-liquid flow dynamics and irregular bubble shapes encountered in gas displacement experiments. To address this challenge, this paper proposes a gas bubble detection and tracking method for gas-driven water microfluidic experiments based on domain adaptation in deep learning and improved YOLOv8. Using the CycleGAN style transfer network to generate numerous simulated bubble training samples significantly reduces the manual labeling workload for bubbles. Additionally, the SE attention mechanism is integrated into the YOLOv8 backbone network to enhance the feature extraction capability of bubbles, and the Fast-PConv Head structure is incorporated into the detection head to enhance detection efficiency. Experimental results demonstrate that using just 20 annotated experimental images, the bubble detection accuracy of the proposed method can reach 94%. Lastly, DeepSORT is employed to track bubbles and visualize the dynamic trends of bubble changes in the form of visual charts. The deep learning approach proposed in this paper provides a new perspective for intelligent analysis of bubbles in microfluidic chip experiments, and future work aiming to its application to the complex multiphase flow field in porous media.