Target Domain Dataset Research Articles

Hyperplane projection network for few-shot relation classification

Recently meta-learning-based few-shot learning methods have been widely used for relation classification. Previous work reveals that meta-learning performs poorly in scenarios where the edge probability distribution of the target domain dataset appears to be significantly different from the source domain. In this paper, we enhance the meta-learning framework with high-dimensional semantic feature extraction and hyperplane projection metrics for meta-tasks. First, we enhance the focus of BERT on entity words by adding entity markers and vector pooling. After that, the high-dimensional semantic features of the support set are extracted and transformed into hyperplanes. Finally, we obtain the classification results by calculating the projection distance between the query sample and the hyperplane. In addition, we design a auxiliary function with a plane correction factor, which can better amplify the plane spacing and reduce the degree of category confusion, which is important for solving the problem of metric spatial loss. Experiments on two real-world few-shot datasets show that our model HPN is more effective in classifying few-shot relations in the same domain and domain-adapted scenarios. And HPN is more stable on NOTA tasks.

Unsupervised cross-domain damage detection and localization for vibration isolators in metro floating-slab track

The steel-spring vibration isolators (SVIs) are critical components of the floating-slab track (FST) in metro system, aiming at ensuring a good performance regarding vibration attenuation. Supervised learning-based models have been recently developed for the SVI damage detection. As they are mostly trained on simulation datasets, the damage detection performance of these models is greatly spoiled on real-life datasets due to the so-called domain shift issue. In this study, a multi-strategy-based domain adaptation methodology is proposed for cross-domain SVI damage detection and localization, where domain adversarial training and feature distribution discrepancy regularization are adopted. The core of the proposed procedure is to extract damage-sensitive and domain-invariant features from the dynamic responses of the track system. The main advantage is that only the source domain dataset needs to be well-labeled while label information of recorded acceleration response related to the real structure is not required for model training. Vehicle-slab track coupled dynamic simulations are conducted to build the labeled source dataset, allowing for different operational scenarios and SVI health conditions. Then, modeling uncertainties are introduced concerning model parameters, and background noise is added to the computed responses to mimic real engineering scenarios in the target domain datasets. A noteworthy good detection performance is reported, when the trained network is finally tested on the target dataset. Ablation studies and feature visualization are then reported, to get insights into the reasons why the proposed method proves superior to a traditional CNN and other domain adaption methods. Finally, an experimental dataset collected during field tests is exploited to validate the effectiveness of the proposed domain adaptation method.

Comparison of synthetic dataset generation methods for medical intervention rooms using medical clothing detection as an example

PurposeThe availability of real data from areas with high privacy requirements, such as the medical intervention space is low and the acquisition complex in terms of data protection. To enable research for assistance systems in the medical intervention room, new methods for data generation for these areas must be researched. Therefore, this work presents a way to create a synthetic dataset for the medical context, using medical clothing object detection as an example. The goal is to close the reality gap between the synthetic and real data.MethodsMethods of 3D-scanned clothing and designed clothing are compared in a Domain-Randomization and Structured-Domain-Randomization scenario using two different rendering engines. Additionally, a Mixed-Reality dataset in front of a greenscreen and a target domain dataset were used while the latter is used to evaluate the different datasets. The experiments conducted are to show whether scanned clothing or designed clothing produce better results in Domain Randomization and Structured Domain Randomization. Likewise, a baseline will be generated using the mixed reality data. In a further experiment it is investigated whether the combination of real, synthetic and mixed reality image data improves the accuracy compared to real data only.ResultsOur experiments show, that Structured-Domain-Randomization of designed clothing together with Mixed-Reality data provide a baseline achieving 72.0% mAP on the test dataset of the clinical target domain. When additionally using 15% (99 images) of available target domain train data, the gap towards 100% (660 images) target domain train data could be nearly closed 80.05% mAP (81.95% mAP). Finally, we show that when additionally using 100% target domain train data the accuracy could be increased to 83.35% mAP.ConclusionIn conclusion, it can be stated that the presented modeling of health professionals is a promising methodology to address the challenge of missing datasets from medical intervention rooms. We will further investigate it on various tasks, like assistance systems, in the medical domain.

A New Defect Diagnosis Method for Wire Rope Based on CNN-Transformer and Transfer Learning

Accurate wire rope defect diagnosis is crucial for the health of whole machinery systems in various industries and practical applications. Although the loss of metallic cross-sectional area signals is the most widely used method in non-destructive wire rope evaluation methods, the weakness and scarcity of defect signals lead to poor diagnostic performance, especially in diverse conditions or those with noise interference. Thus, a new wire rope defect diagnosis method is proposed in this study. First, empirical mode decomposition and isolation forest methods are applied to eliminate noise signals and to locate the defects. Second, a convolution neural network and transformer encoder are used to design a new wire rope defect diagnosis network for the improvement of the feature extraction ability. Third, transfer learning architecture is established based on gray feature images to fine-tune the pre-trained model using a small target domain dataset. Finally, comparison experiments and a visualization analysis are conducted to verify the effectiveness of the proposed methods. The results demonstrate that the presented model can improve the performance of the wire rope defect diagnosis method under cross-domain conditions. Additionally, the transfer feasibility of transfer learning architecture is discussed for future practical applications.

The accuracy losing phenomenon in abrasive tool condition monitoring and a noval WMMC-JDA based data-driven method considered tool stochastic surface morphology

As a critically issue in the maintaining of desired part quality and manufacturing productivity, tool condition monitoring (TCM) problem is treated by multi data-driven machine learning methods with training and testing datasets generated by stochastic segmentation. For existing methods, it is assumed that samples collected from same type tools follow identical probability distribution. Different from regular tools, abrasive tools dissatisfy this assumption because of their stochastic surface morphologies. Therefore, traditional data-driven methods loss their accuracies in abrasive tool changing cases. In data driven TCM models, the macroscopic effect of abrasive tool micro-stochasticity is discovered. For the abrasive tool frequently changing cases, an abrasive tool condition monitoring (ATCM) method based on unsupervised domain adaptation is presented to deal with changes caused by stochastic surface morphology. In feature-based transfer learning, source and target domain datasets are projected into a transferred subspace to enhance the distribution alignment similarity of two domains. A weighted maximum margin criterion (WMMC) is adopted in the domain adaptation process to make the transformed samples in same class closing but segregated from those in different classes. With unsupervised domain adaptation mechanism and multilayer perceptron (MLP), the unsupervised domain adaptation is proposed to deal with traditional model invalidation after abrasive tool changes. The proposed method is verified by the robotic belt grinding experiment results. Compared with the existing data-driven method, the weighted maximum margin criterion joint distribution adaptation (WMMC-JDA) based ATCM method still maintains the predicting effectiveness after abrasive tool changes.

Deep transfer learning strategy for efficient domain generalisation in machine fault diagnosis

Automated fault diagnosis algorithms based on vibration sensor recordings play an important role in determining the state of health of the machines. Data-driven approaches demand a large amount of labelled data to build reliable models. The performance of such lab-trained models degrades when deployed in practical use cases in the presence of distinct distribution target domain datasets. In this work, we present a novel deep transfer learning strategy that fine-tunes the trainable parameters of the lower (convolutional) layers with respect to the changing target domain datasets and transfers the parameters of the deeper (dense) layers from the source domain for efficient domain generalisation and fault classification. The performance of this strategy is evaluated by considering two different target domain datasets and studying the sensitivity of fine-tuning individual layers in the networks using time-frequency representations of the vibration signals (scalograms) as inputs. We observe that the proposed transfer learning strategy yields near-perfect accuracy, even for use cases where low-precision sensors are used for data collection and unlabelled run-to-failure data with a limited number of training samples.

An instance-based deep transfer learning method for quality identification of Longjing tea from multiple geographical origins

For practitioners, it is very crucial to realize accurate and automatic vision-based quality identification of Longjing tea. Due to the high similarity between classes, the classification accuracy of traditional image processing combined with machine learning algorithm is not satisfactory. High-performance deep learning methods require large amounts of annotated data, but collecting and labeling massive amounts of data is very time consuming and monotonous. To gain as much useful knowledge as possible from related tasks, an instance-based deep transfer learning method for the quality identification of Longjing tea is proposed. The method mainly consists of two steps: (i) The MobileNet V2 model is trained using the hybrid training dataset containing all labeled samples from source and target domains. The trained MobileNet V2 model is used as a feature extractor, and (ii) the extracted features are input into the proposed multiclass TrAdaBoost algorithm for training and identification. Longjing tea images from three geographical origins, West Lake, Qiantang, and Yuezhou, are collected, and the tea from each geographical origin contains four grades. The Longjing tea from West Lake is regarded as the source domain, which contains more labeled samples. The Longjing tea from the other two geographical origins contains only limited labeled samples, which are regarded as the target domain. Comparative experimental results show that the method with the best performance is the MobileNet V2 feature extractor trained with a hybrid training dataset combined with multiclass TrAdaBoost with linear support vector machine (SVM). The overall Longjing tea quality identification accuracy is 93.6% and 91.5% on the two target domain datasets, respectively. The proposed method can achieve accurate quality identification of Longjing tea with limited samples. It can provide some heuristics for designing image-based tea quality identification systems.

Passive Thermography Based Bearing Fault Diagnosis Using Transfer Learning With Varying Working Conditions

Bearing is one of the core components of any rotating machine, and its failure is widespread. This reason drives continuous monitoring and detecting bearing faults during machine operation to warn operators and prevent unforeseen damage. This paper proposes an intelligent Passive Thermography (PTG) based fault diagnosis technique for detection of bearing faults using Convolutional Neural Network (CNN) with Transfer Learning (TL) under varying working conditions. The validation of the proposed method has been done on three different datasets; bearing test rig dataset has been taken as a source domain data while Induction Motor (IM), and Machine Fault Simulator (MFS) datasets have been taken as target domain data. The proposed method enables and speeds up the training process of CNN towards accurate adaptation for fault diagnosis approach in the escalated time frame. The experimental results demonstrated that the suggested approach could successfully learn transferable characteristics from the source domain model, which can cope with the issue of limited availability of training data required for the target domain. The classification accuracy on the target domain dataset were varied in the range of 89 -95.4 % in the case of the IM dataset and 96.5-97.5% in the case of the MFS dataset. Moreover, it shows the benefits of the suggested method, which may be utilized as an effective non-invasive diagnostic tool for rotating machines to avoid unexpected system shutdowns.

Source free unsupervised domain adaptation for electro-mechanical actuator fault diagnosis

A common necessity for prior unsupervised domain adaptation methods that can improve the domain adaptation in unlabeled target domain dataset is access to source domain dataset and target domain dataset simultaneously. However, data privacy makes it not always possible to access source domain dataset and target domain dataset in actual industrial equipment simultaneously, especially for aviation component like Electro-Mechanical Actuator (EMA) whose dataset are often not shareable due to the data copyright and confidentiality. To address this problem, this paper proposes a source free unsupervised domain adaptation framework for EMA fault diagnosis. The proposed framework is a combination of feature network and classifier. Firstly, source domain datasets are only applied to train a source model. Secondly, the well-trained source model is transferred to target domain and classifier is frozen based on source domain hypothesis. Thirdly, nearest centroid filtering is introduced to filter the reliable pseudo labels for unlabeled target domain dataset, and finally, supervised learning and pseudo label clustering are applied to fine-tune the transferred model. In comparison with several traditional unsupervised domain adaptation methods, case studies based on low- and high-frequency monitoring signals on EMA indicate the effectiveness of the proposed method.

EasyDAM_V3: Automatic Fruit Labeling Based on Optimal Source Domain Selection and Data Synthesis via a Knowledge Graph.

Although deep learning-based fruit detection techniques are becoming popular, they require a large number of labeled datasets to support model training. Moreover, the manual labeling process is time-consuming and labor-intensive. We previously implemented a generative adversarial network-based method to reduce labeling costs. However, it does not consider fitness among more species. Methods of selecting the most suitable source domain dataset based on the fruit datasets of the target domain remain to be investigated. Moreover, current automatic labeling technology still requires manual labeling of the source domain dataset and cannot completely eliminate manual processes. Therefore, an improved EasyDAM_V3 model was proposed in this study as an automatic labeling method for additional classes of fruit. This study proposes both an optimal source domain establishment method based on a multidimensional spatial feature model to select the most suitable source domain, and a high-volume dataset construction method based on transparent background fruit image translation by constructing a knowledge graph of orchard scene hierarchy component synthesis rules. The EasyDAM_V3 model can automatically obtain fruit label information from the dataset, thereby eliminating manual labeling. To test the proposed method, pear was used as the selected optimal source domain, followed by orange, apple, and tomato as the target domain datasets. The results showed that the average precision of annotation reached 90.94%, 89.78%, and 90.84% for the target datasets, respectively. The EasyDAM_V3 model can obtain the optimal source domain in automatic labeling tasks, thus eliminating the manual labeling process and reducing associated costs and labor.

Source-Free Domain Adaptive Detection of Concealed Objects in Passive Millimeter-Wave Images

Passive millimeter-wave (PMMW) imager can detect concealed objects under clothing in a touch-free manner. Existing detectors exhibit promising performance on specific PMMW datasets. However, these detectors will suffer severe degradation due to the inevitable domain shift when the environment changes in real applications. To avoid huge manual efforts in preparing the target domain dataset, unsupervised domain adaption (UDA) is proposed to tackle this issue. UDA holds the assumption that the labeled source data and unlabeled target data are accessible during the adaptation, but it is not always true due to the privacy concern. In this article, we aim to address this challenge by source-free domain adaption (SFDA), which migrates the source-trained model to target domain without the source data. Our method is built on the student–teacher framework and combines the merits of pseudo-label self-training and adversarial learning (AL). We also propose a novel spatio-temporal weighting strategy (STW). It mines the spatial self-correlation in single PMMW image and temporal cross correlation among PMMW images for teacher to produce high-quality pseudo-labels. It guides the student to select strong positive pairs for contrastive learning. In addition, the AL is further employed for feature alignment between teacher and student at image and instance levels, which ensures the student to learn domain-invariant features. Extensive experiments show that our method achieves the state-of-the-art performance. We further demonstrate that the proposed method can work in online settings. To the best of authors’ knowledge, this is the first work to incorporate SFDA for concealed object detection in PMMW images.

Unsupervised domain adaptation using modified cycle generative adversarial network for aerial image classification

The unsupervised domain adaptation (UDA) of aerial image classification is of great significance in the traffic management and public safety applications. Due to the differences in the distribution of source and target domain datasets from different sensors and cities, and the acquisition difficulty and high annotation cost of target domain datasets, existing domain adaptive methods have difficulty obtaining high accuracy and stability. In this work, a modified cycle generative adversarial network (CycleGAN) (MGAN) based on the multiscale residual block (MSRB) and multipooling coordinate attention (MPCA) modules is proposed to generate fake images, which is the most critical stage of the UDA method. In the MGAN, the MSRB module can extract the multiscale information of the image, and the MPCA module can fuse the position and channel information in the image. These two modules better extract the features of edges, regions and semantic classes, which makes the generated fake images better retain the information of the source domain and possess the style of the target domain. Otherwise, the Deeplabv3+ model is used to obtain a pretrained and final segmentation model in the first and third stages. In the experiment, the proposed method is validated on the aerial images of the Potsdam and Vaihingen datasets in the International Society for Photogrammetry and Remote Sensing (ISPRS) two-dimensional semantic labeling benchmark and improves overall accuracy from 52% to 60% on the segmentation of the target domain, compared with the existing optimal method; it also achieves an accuracy of 77% for buildings. The results demonstrate that the proposed method can achieve a higher accuracy and stability.

Continual BatchNorm Adaptation (CBNA) for Semantic Segmentation

Environment perception in autonomous driving vehicles often heavily relies on deep neural networks (DNNs), which are subject to domain shifts, leading to a significantly decreased performance during DNN deployment. Usually, this problem is addressed by unsupervised domain adaptation (UDA) approaches trained either simultaneously on source and target domain datasets or even source-free only on target data in an offline fashion. In this work, we further expand a source-free UDA approach to a continual and therefore online-capable UDA on a single-image basis for semantic segmentation. Accordingly, our method only requires the pre-trained model from the supplier (trained in the source domain) and the current (unlabeled target domain) camera image. Our method Continual BatchNorm Adaptation (CBNA) modifies the source domain statistics in the batch normalization layers, using target domain images in an unsupervised fashion, which yields consistent performance improvements during inference. Thereby, in contrast to existing works, our approach can be applied to improve a DNN continuously on a single-image basis during deployment without access to source data, without algorithmic delay, and nearly without computational overhead. We show the consistent effectiveness of our method across a wide variety of source/target domain settings for semantic segmentation. Code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/ifnspaml/CBNA</uri>

TinyML-enabled edge implementation of transfer learning framework for domain generalization in machine fault diagnosis

TinyML has the potential to be a huge enabler of smart sensor nodes for fault diagnosis of machines by embedding powerful machine learning algorithms in low-cost edge devices. Static models trained with pre-collected data may not perform satisfactorily when deployed in use cases that present data with a significantly different distribution, due to changes in machine instances, installation, and environment and operating conditions. In this work, we present a transfer learning framework in conjunction with TinyML-powered convolutional neural network (CNN) architecture for vibration-based fault diagnosis of vastly different machines. We report that fine-tuning (retraining) of parameters (weights and biases) in convolutional layers and transfer of parameters in dense layers of CNNs results in higher accuracy for domain generalization than the conventional approach of transferring parameters for convolutional layers and fine-tuning dense layers. Using transfer learning also results in high accuracy when the total number of training samples used for fine-tuning on the target domain data sets are as low as 25 i.e. less than 5% of the size of the data set, which makes the proposed architecture viable for scenarios with a limited number of data points available for training the model. Furthermore, we have also demonstrated memory-efficient retraining at the edge by fine-tuning only the biases of hidden layers.

Feature Extraction Method of Art Visual Communication Image Based on 5G Intelligent Sensor Network

5G intelligent sensor network technology realizes the perception, processing, and transmission of information. It forms the three pillars of information technology together with computer technology and communication technology and is an important part of the Internet of Things technology. The 5G smart sensor network is a wireless communication module added to the sensor nodes, and a wireless communication network is formed by a large number of stationary or movable sensor nodes in the form of self-organization and multihop transmission. This paper proposes a keypoint feature extraction method based on deep learning, which can extract keypoint local features for matching. This method uses the convolutional network structure, which is pretrained based on the Siamese network structure and then adjusted to the ternary network structure to continue training to improve the accuracy. This paper proposes a high-art visual communication image classification based on multifeature extraction and classification decision fusion. In the data preprocessing stage, the correlation alignment algorithm is performed on the datasets of different domains (source domain and target domain) to reduce the difference in spatial distribution, and then, a multifeature extractor is designed to extract artistic visual communication images and spatial information. In the process, the multitask learning method is introduced to jointly train the networks of multiple data sets to reduce the degree of overfitting of the model, solve the problem of insufficient labeled samples in the target domain data set, and affect the classification accuracy of high-art visual communication images. Finally, the classification results are obtained through the fusion of voting decisions. The experimental results show that the advantage of this framework is that it utilizes the artistic visual communication image and spatial structure information from the source and target scenes, which can significantly reduce the dependence on the number of labeled samples in the target domain and improve the classification performance. In this paper, a dual-channel deep residual convolutional neural network is designed. The multiple convolution layers of the residual module in the network use hard parameters to share, so that the deep feature representation on the joint spatial spectrum dimension can be automatically extracted. The features extracted by the network are transferred to maximize the auxiliary role of the labeled samples in the source domain and avoid the negative transfer problem caused by the forced transfer between irrelevant samples.

H-net: Unsupervised domain adaptation person re-identification network based on hierarchy

Due to the high cost of manual labeling for supervised person re-identification (re-ID), unsupervised domain adaptation (UDA) person re-ID has been attracting the attention of many scholars. In this research, target domain datasets and source domain datasets are two indispensable datasets, and although there are many different pictures of the same person in the target domain, these pictures are precious to the network in different degrees. However, the existing UDA person re-ID algorithms does not treat different samples in the target domain differently, they just treat positive samples as indistinguishable samples. Not only that, although the triplet loss has been re-identified by unsupervised person re-ID, the noise of the hardest sample hasn't been carried out well. In this paper, a novel and robust network model named unsupervised domain adaptation hierarchical person re-identification network (H-Net) is proposed, which not only effectively reduces the impact of inaccurate identification of the hardest sample but also treats different positive samples differently by hierarchical feature collection. Numerous experimental results on Market-1501 and DukeMTMC-reID demonstrate that the proposed H-Net outperforms the existing methods and can significantly improve the accuracy of person re-ID.

Deep-Learning-Based Financial Message Sentiment Classification in Business Management.

A deep-learning-based financial text sentiment classification method is proposed in this paper, which can provide a reference for business management. In the proposed method, domain adaptation is adopted to solve the common problem of insufficient labeled samples in the financial textual domain. Specifically, in the classification process, the seq2seq model is firstly adopted to extract the abstract from the financial message, which can reduce the influence of invalid information and speed up processing. In the process of sentiment classification, a bidirectional LSTM model is adopted for classification, which can more comprehensively make use of context information. Experiments are carried out to testify the proposed method through the open-source data set. It can be seen that the proposed method can effectively transfer from the reduced Amazon data set to the StockTwits financial text data set. Compared with the parameter-frozen-based method and the SDA-based method, the recognition rates have improved by 0.5% and 6.8%, respectively. If the target domain data set can be directly adopted for training, the recognition rate of the proposed method is higher than that of the SVM method and the LSTM method by 8.3% and 4.5%, respectively.

Transfer Learning-Based Artificial Neural Network for Forward Kinematic Estimation of 6-DOF Robot

Transfer Learning (TL) can significantly lower training time and reduce dependency on a large number of target domain datasets. Such an approach is still not exploited for robotic prediction tasks. Currently, a TL based Artificial Neural Network (ANN) is explored and validated for robotic forward kinematics estimation of a 6-DOF robot. The robotic positions are estimated from the available joint angle information. The 6-R MTAB Aristo-XT robot is selected as a case study to generate the target experimental training and testing data for validation of ML techniques. While, the PUMA 560 6-DOF robot is used as a source for prior training of the ANN model. Standard performance measures such as learning error, deviation error and Mean Square Error (MSE) are evaluated and graphical illustrations are presented for fair comparison of the results. Experimental results reveal that, instead of ANN, the TL-ANN is strongly suggested to improve the training time of ANN regressor, and it also reduces the randomness and improves the accuracy as compared to its counterpart.

Deep Learning Domain Adaptation for Electro-Mechanical Actuator Fault Diagnosis Under Variable Driving Waveforms

Benefited from the sampling signal from monitoring sensors installed in electro-mechanical actuator (EMA), data-driven methods, such as deep learning models that can mine adaptive features from monitoring signal, have received wide attention in EMA fault diagnosis. A generalized hypothesis accepted by previous efforts on EMA fault diagnosis, is that target domain datasets (i.e., testing monitoring signal) are drawn from the same distribution with source domain datasets (i.e., training monitoring signal). However, except for the different working loads and velocities, various driving signal waveforms also exits in EMA, which further increase the domain discrepancy challenge and decrease the performance of those deep learning models. Therefore, this paper proposes a deep learning domain adaptation method that introduces joint maximum mean discrepancy (JMMD) into convolutional neural network (CNN), which aims to reduce the domain difference between labeled source domain datasets and unlabeled target domain datasets. In order to simulate domain discrepancy phenomenon, the NASA EMA datasets, sampled from acceleration sensors under variable driving waveforms with different working loads and velocities, are considered in this paper. Experiment results indicate that CNN-JMMD method can reduce the distribution difference and improve the diagnosis accuracy in source and target domains. In comparison with several available methods, extensive experimental evaluations on six domain tasks indicate the effectiveness of the proposed method. In addition, better performance has been obtained by selecting the proper trade-off factor between CNN and JMMD.

Transfer learning with molecular graph convolutional networks for accurate modeling and representation of bioactivities of ligands targeting GPCRs without sufficient data

There are many new or potential drug targets in G protein-coupled receptors (GPCRs) without sufficient ligand associations, and it is essential and urgent to implement drug discovery targeting these GPCRs. Precise modeling and representing ligand bioactivities are essential for screening and optimizing these GPCR targeted drugs, yet insufficient samples made it difficult to achieve. Transfer learning intends to solve this by introducing rich information from related source domains with sufficient ligand training samples. In addition, ligand molecules naturally constitute a graph structure, which can be utilized by molecular graph convolutional networks to form an end-to-end multiple-level representation learning. This study proposed a novel method, TL-MGCN, using transfer learning with molecular graph convolutional networks to precisely model and represent bioactivities of ligands targeting GPCRs without sufficient data. The study tested TL-MGCN on a series of 54 representative target domain datasets which cover most human subfamilies, and 44 out of them have less than 600 ligand associations. TL-MGCN obtained an average improvement of 28.74%, 17.28%, 10.05%, 77.83%, 43.65% and 14.65% on correlation coefficient (r2) and 11.90%, 7.43%, 14.86%, 41.46%, 31.02% and 22.94% on root-mean-square error (RMSE) compared with the WDL-RF, transfer learning version of WDL-RF (TR-WDL-RF), attentive FP, GIN, Weave and MPNN predictors, respectively. Users have free access to the web server of TL-MGCN, along with the source codes and datasets, at http://www.noveldelta.com/TL_MGCN for academic purposes.