Transfer Learning Approach Research Articles

Low Resource Malay Dialect Automatic Speech Recognition Modeling Using Transfer Learning from a Standard Malay Model

Approaches to automatic speech recognition have transited from Hidden Markov Model (HMM)-based ASR to deep neural networks. The advantages of deep neural network approaches are that they can be developed quickly and perform better given large language resources. Nevertheless, dialect speech recognition is still challenging due to the limited resources. Transfer learning approaches have been proposed to improve speech recognition for low resources. In the first approach, the model is pre-trained on a large and diverse labeled dataset to learn the acoustic and language patterns from the speech signal. Then, the model parameters are updated with a new dataset, and the pre-trained model is fine-tuned on a low-resource language dataset. The fine-tuning process is usually completed by freezing the pre-trained layers and training the remaining layers of the model on the low-resource language corpus. Another approach is to use a pre-trained model to capture the compact and meaningful features as input to the encoder. Pre-training in this approach usually involves using unsupervised learning methods to train models on a corpus of large amounts of unmarked data. It enables the model to learn the general patterns and relationships between the input speech signals. This paper proposes a training recipe using transfer learning and Standard Malay models to improve automatic speech recognition for Kelantan and Sarawak Malay dialects.

Ensemble approach of transfer learning and vision transformer leveraging explainable AI for disease diagnosis: An advancement towards smart healthcare 5.0

Smart healthcare has advanced the medical industry with the integration of data-driven approaches. Artificial intelligence and machine learning provided remarkable progress, but there is a lack of transparency and interpretability in such applications. To overcome such limitations, explainable AI (EXAI) provided a promising result. This paper applied the EXAI for disease diagnosis in the advancement of smart healthcare. The paper combined the approach of transfer learning, vision transformer, and explainable AI and designed an ensemble approach for prediction of disease and its severity. The result is evaluated on a dataset of Alzheimer's disease. The result analysis compared the performance of transfer learning models with the ensemble model of transfer learning and vision transformer. For training, InceptionV3, VGG19, Resnet50, and Densenet121 transfer learning models were selected for ensembling with vision transformer. The result compares the performance of two models: a transfer learning (TL) model and an ensemble transfer learning (Ensemble TL) model combined with vision transformer (ViT) on ADNI dataset. For the TL model, the accuracy is 58 %, precision is 52 %, recall is 42 %, and the F1-score is 44 %. Whereas, the Ensemble TL model with ViT shows significantly improved performance i.e., 96 % of accuracy, 94 % of precision, 90 % of recall and 92 % of F1-score on ADNI dataset. This shows the efficacy of the ensemble model over transfer learning models.

A Transfer Learning Approach: Early Prediction of Alzheimer’s Disease on US Healthy Aging Dataset

Alzheimer’s disease (AD) is a growing public health crisis, a very global health concern, and an irreversible progressive neurodegenerative disorder of the brain for which there is still no cure. Globally, it accounts for 60–80% of dementia cases, thereby raising the need for an accurate and effective early classification. The proposed work used a healthy aging dataset from the USA and focused on three transfer learning approaches: VGG16, VGG19, and Alex Net. This work leveraged how the convolutional model and pooling layers work to improve and reduce overfitting, despite challenges in training the numerical dataset. VGG was preferably chosen as a hidden layer as it has a more diverse, deeper, and simpler architecture with better performance when dealing with larger datasets. It consumes less memory and training time. A comparative analysis was performed using machine learning and neural network algorithm techniques. Performance metrics such as accuracy, error rate, precision, recall, F1 score, sensitivity, specificity, kappa statistics, ROC, and RMSE were experimented with and compared. The accuracy was 100% for VGG16 and VGG19 and 98.20% for Alex Net. The precision was 99.9% for VGG16, 96.6% for VGG19, and 100% for Alex Net; the recall values were 99.9% for all three cases of VGG16, VGG19, and Alex Net; and the sensitivity metric was 96.8% for VGG16, 97.9% for VGG19, and 98.7% for Alex Net, which has outperformed when compared with the existing approaches for the classification of Alzheimer’s disease. This research contributes to the advancement of predictive knowledge, leading to future empirical evaluation, experimentation, and testing in the biomedical field.

Sparse coding-based transfer learning for energy disaggregation

Energy disaggregation, or Non-Intrusive Load Monitoring (NILM), is a technique that predicts the consumption levels of individual appliances from only the main signal in the building. Various methods have been proposed to solve this problem, including sparse coding (SC), which offers great advantages due to its ability to capture complex patterns in data. However, a challenging aspect of NILM is that data containing appliance-level information is scarce. Moreover, the houses that the models are tested on might be from a different population than the training data, thus resulting in a domain shift. Therefore, we need to develop approaches that are adapted to training data scarcity through the use of transfer learning (TL). In this paper, we test several TL approaches on SC models with the aim of discriminative energy disaggregation (DD). We propose four TL SC models: “Transfer SC+DD”, “SC source, DD target”, “Transfer Deep SC” and “Transfer Variational SC”. We compare these models with 4 models that do not apply domain adaptation. We show that TL-based models achieve similar and sometimes better results than the regular models. We also show that further training on only the aggregate signal of some target domain houses improves the performance of a SC model trained only on the source.

Pulmonary nodule detection in low dose computed tomography using a medical-to-medical transfer learning approach.

Lung cancer is the second most common cancer and the leading cause of cancer death globally. Low dose computed tomography (LDCT) is the recommended imaging screening tool for the early detection of lung cancer. A fully automated computer-aided detection method for LDCT will greatly improve the existing clinical workflow. Most of the existing methods for lung detection are designed for high-dose CTs (HDCTs), and those methods cannot be directly applied to LDCTs due to domain shifts and inferior quality of LDCT images. In this work, we describe a semi-automated transfer learning-based approach for the early detection of lung nodules using LDCTs. In this work, we developed an algorithm based on the object detection model, you only look once (YOLO) to detect lung nodules. The YOLO model was first trained on CTs, and the pre-trained weights were used as initial weights during the retraining of the model on LDCTs using a medical-to-medical transfer learning approach. The dataset for this study was from a screening trial consisting of LDCTs acquired from 50 biopsy-confirmed lung cancer patients obtained over 3 consecutive years (T1, T2, and T3). About 60 lung cancer patients' HDCTs were obtained from a public dataset. The developed model was evaluated using a hold-out test set comprising 15 patient cases (93 slices with cancerous nodules) using precision, specificity, recall, and F1-score. The evaluation metrics were reported patient-wise on a per-year basis and averaged for 3 years. For comparative analysis, the proposed detection model was trained using pre-trained weights from the COCO dataset as the initial weights. A paired t-test and chi-squared test with an alpha value of 0.05 were used for statistical significance testing. The results were reported by comparing the proposed model developed using HDCT pre-trained weights with COCO pre-trained weights. The former approach versus the latter approach obtained a precision of 0.982 versus 0.93 in detecting cancerous nodules, specificity of 0.923 versus 0.849 in identifying slices with no cancerous nodules, recall of 0.87 versus 0.886, and F1-score of 0.924 versus 0.903. As the nodule progressed, the former approach achieved a precision of 1, specificity of 0.92, and sensitivity of 0.930. The statistical analysis performed in the comparative study resulted in a -value of 0.0054 for precision and a -value of 0.00034 for specificity. In this study, a semi-automated method was developed to detect lung nodules in LDCTs using HDCT pre-trained weights as the initial weights and retraining the model. Further, the results were compared by replacing HDCT pre-trained weights in the above approach with COCO pre-trained weights. The proposed method may identify early lung nodules during the screening program, reduce overdiagnosis and follow-ups due to misdiagnosis in LDCTs, start treatment options in the affected patients, and lower the mortality rate.

Vision transformer and CNN-based skin lesion analysis: classification of monkeypox

AbstractMonkeypox is an important health problem. Rapid diagnosis of monkeypox skin lesions and emergency isolation when necessary is essential. Also, some skin lesions, such as melanoma, can be fatal and must be rapidly distinguished. However, in some cases, it is difficult to distinguish the lesions visually. Methods such as dermoscopy, high-resolution ultrasound imaging, etc. can be used for better observation. But these methods are often based on qualitative analysis, subjective and time-consuming. Therefore, in this study, a quantitative and objective classification tool has been developed to assist dermatologists and scientists. The proposed system classifies seven skin lesions, including monkeypox. A popular approach Vision Transformer and some popular deep learning convolutional networks have been trained with the transfer learning approach and all results have been compared. Then, the models that show the best accuracy score have been combined to make the final prediction using bagging-ensemble learning. The proposed ensemble-based system produced 81.91% Accuracy, 65.94% Jaccard, 87.16% Precision, 74.12% Recall, and 78.16% Fscore values. In terms of different criteria metrics, the system produced competitive or even better results than the literature.

Efficient deep learning-based automated diagnosis from echocardiography with contrastive self-supervised learning

BackgroundAdvances in self-supervised learning (SSL) have enabled state-of-the-art automated medical image diagnosis from small, labeled datasets. This label efficiency is often desirable, given the difficulty of obtaining expert labels for medical image recognition tasks. However, most efforts toward SSL in medical imaging are not adapted to video-based modalities, such as echocardiography.MethodsWe developed a self-supervised contrastive learning approach, EchoCLR, for echocardiogram videos with the goal of learning strong representations for efficient fine-tuning on downstream cardiac disease diagnosis. EchoCLR pretraining involves (i) contrastive learning, where the model is trained to identify distinct videos of the same patient, and (ii) frame reordering, where the model is trained to predict the correct of video frames after being randomly shuffled.ResultsWhen fine-tuned on small portions of labeled data, EchoCLR pretraining significantly improves classification performance for left ventricular hypertrophy (LVH) and aortic stenosis (AS) over other transfer learning and SSL approaches across internal and external test sets. When fine-tuning on 10% of available training data (519 studies), an EchoCLR-pretrained model achieves 0.72 AUROC (95% CI: [0.69, 0.75]) on LVH classification, compared to 0.61 AUROC (95% CI: [0.57, 0.64]) with a standard transfer learning approach. Similarly, using 1% of available training data (53 studies), EchoCLR pretraining achieves 0.82 AUROC (95% CI: [0.79, 0.84]) on severe AS classification, compared to 0.61 AUROC (95% CI: [0.58, 0.65]) with transfer learning.ConclusionsEchoCLR is unique in its ability to learn representations of echocardiogram videos and demonstrates that SSL can enable label-efficient disease classification from small amounts of labeled data.

PowerTrain: Fast, generalizable time and power prediction models to optimize DNN training on accelerated edges

Accelerated edge devices, like Nvidia’s Jetson with 1000+ CUDA cores, are increasingly used for DNN training and federated learning, rather than just for inferencing workloads. A unique feature of these compact devices is their fine-grained control over CPU, GPU, memory frequencies, and active CPU cores, which can limit their power envelope in a constrained setting while throttling the compute performance. Given this vast 10k+ parameter space, selecting a power mode for dynamically arriving training workloads to exploit power–performance trade-offs requires costly profiling for each new workload, or is done ad hoc. We propose PowerTrain, a transfer-learning approach to accurately predict the power and time that will be consumed when we train a given DNN workload (model + dataset) using any specified power mode (CPU/GPU/memory frequencies, core-count). It requires a one-time offline profiling of 1000s of power modes for a reference DNN workload on a single Jetson device (Orin AGX) to build Neural Network (NN) based prediction models for time and power. These NN models are subsequently transferred (retrained) for a new DNN workload, or even a different Jetson device, with minimal additional profiling of just 50 power modes to make accurate time and power predictions. These are then used to rapidly construct the Pareto front and select the optimal power mode for the new workload, e.g., to minimize training time while meeting a power limit. PowerTrain’s predictions are robust to new workloads, exhibiting a low MAPE of <6% for power and <15% for time on six new training workloads (MobileNet, YOLO, BERT, LSTM, etc.) for up to 4400 power modes, when transferred from a ResNet reference workload on Orin AGX. It is also resilient when transferred to two entirely new Jetson devices (Xavier AGX and Jetson Orin Nano) with prediction errors of <14.5% and <11%. These outperform baseline predictions by more than 10% and baseline optimizations by up to 45% on time and 88% on power.

A Comprehensive Review on Machine Learning and Transfer Learning Approaches in Liver Tumor Classification

A Comprehensive Review on Machine Learning and Transfer Learning Approaches in Liver Tumor Classification

EarlyNet: a novel transfer learning approach with VGG11 and EfficientNet for early-stage breast cancer detection

EarlyNet: a novel transfer learning approach with VGG11 and EfficientNet for early-stage breast cancer detection

Predicting the Prognosis of HIFU Ablation of Uterine Fibroids Using a Deep Learning-Based 3D Super-Resolution DWI Radiomics Model: A Multicenter Study

Rationale and ObjectivesTo assess the feasibility and efficacy of a deep learning based three-dimensional (3D) super-resolution diffusion-weighted imaging (DWI) radiomics model in predicting the prognosis of high-intensity focused ultrasound (HIFU) ablation of uterine fibroids. MethodsThis retrospective study included 360 patients with uterine fibroids who received HIFU treatment, including Center A (training set: N=240; internal testing set: N=60) and Center B (external testing set: N=60) and were classified as having a favorable or unfavorable prognosis based on the postoperative non-perfusion volume ratio. A deep transfer learning approach was used to construct super-resolution DWI (SR-DWI) based on conventional high-resolution DWI (HR-DWI), and 1198 radiomics features were extracted from manually segmented regions of interest in both image types. Following data preprocessing and feature selection, radiomics models were constructed for HR-DWI and SR-DWI using Support Vector Machine (SVM), Random Forest (RF), and Light Gradient Boosting Machine (LightGBM) algorithms, with performance evaluated using area under the curve (AUC) and decision curves. ResultAll DWI radiomics models demonstrated superior AUC in predicting HIFU ablated uterine fibroids prognosis compared to expert radiologists (AUC: 0.706, 95% CI: 0.647-0.748). When utilizing different machine learning algorithms, the HR-DWI model achieved AUC values of 0.805 (95% CI: 0.679-0.931) with SVM, 0.797 (95% CI: 0.672-0.921) with RF, and 0.770 (95% CI: 0.631-0.908) with LightGBM. Meanwhile, the SR-DWI model outperformed the HR-DWI model (P<0.05) across all algorithms, with AUC values of 0.868 (95% CI: 0.775-0.960) with SVM, 0.824 (95% CI: 0.715-0.934) with RF, and 0.821 (95% CI: 0.709-0.933) with LightGBM. And decision curve analysis further confirmed the good clinical value of the models. ConclusionDeep learning based 3D SR-DWI radiomics model demonstrated favorable feasibility and effectiveness in predicting the prognosis of HIFU ablated uterine fibroids, which was superior to HR-DWI model and assessment by expert radiologists.

Application of Deep Learning for Segmenting Seepages in Levee Systems

Seepage is a typical hydraulic factor that can initiate the breaching process in a levee system. If not identified and treated on time, seepages can be a severe problem for levees, weakening the levee structure and eventually leading to collapse. Therefore, it is essential always to be vigilant with regular monitoring procedures to identify seepages throughout these levee systems and perform adequate repairs to limit potential threats from unforeseen levee failures. This paper introduces a fully convolutional neural network to identify and segment seepage from the image in levee systems. To the best of our knowledge, this is the first work in this domain. Applying deep learning techniques for semantic segmentation tasks in real-world scenarios has its own challenges, especially the difficulty for models to effectively learn from complex backgrounds while focusing on simpler objects of interest. This challenge is particularly evident in the task of detecting seepages in levee systems, where the fault is relatively simple compared to the complex and varied background. We addressed this problem by introducing negative images and a controlled transfer learning approach for semantic segmentation for accurate seepage segmentation in levee systems.

Deep mutational scanning reveals sequence to function constraints for SWEET family transporters.

Protein science is entering a transformative phase enabled by deep mutational scans that provide an unbiased view of the residue level interactions that mediate function. However, it has yet to be extensively used to characterize the mutational and evolutionary landscapes of plant proteins. Here, we apply the method to explore sequence-function relationships within the sugar transporter AtSWEET13. DMS results describe how mutational interrogation throughout different regions of the protein affects AtSWEET13 abundance and transport function. Our results identify novel transport-enhancing mutations that are validated using the FRET sensor assays. Extending DMS results to phylogenetic analyses reveal the role of transmembrane helix 4 (TM4) which makes the SWEET family transporters distinct from prokaryotic SemiSWEETs. We show that transmembrane helix 4 is intolerant to motif swapping with other clade-specific SWEET TM4 compositions, despite accommodating single point-mutations towards aromatic and charged polar amino acids. We further show that the transfer learning approaches based on physics and ML based In silico variant prediction tools have limited utility for engineering plant proteins as they were unable to reproduce our experimental results. We conclude that DMS can produce datasets which, when combined with the right predictive computational frameworks, can direct plant engineering efforts through derivative phenotype selection and evolutionary insights.

Heterogeneous latent transfer learning in Gaussian graphical models.

Gaussian graphical models (GGMs) are useful for understanding the complex relationships between biological entities. Transfer learning can improve the estimation of GGMs in a target dataset by incorporating relevant information from related source studies. However, biomedical research often involves intrinsic and latent heterogeneity within a study, such as heterogeneous subpopulations. This heterogeneity can make it difficult to identify informative source studies or lead to negative transfer if the source study is improperly used. To address this challenge, we developed a heterogeneous latent transfer learning (Latent-TL) approach that accounts for both within-sample and between-sample heterogeneity. The idea behind this approach is to "learn from the alike" by leveraging the similarities between source and target GGMs within each subpopulation. The Latent-TL algorithm simultaneously identifies common subpopulation structures among samples and facilitates the learning of target GGMs using source samples from the same subpopulation. Through extensive simulations and real data application, we have shown that the proposed method outperforms single-site learning and standard transfer learning that ignores the latent structures. We have also demonstrated the applicability of the proposed algorithm in characterizing gene co-expression networks in breast cancer patients, where the inferred genetic networks identified many biologically meaningful gene-gene interactions.

Using autoencoders and deep transfer learning to determine the stellar parameters of 286 CARMENES M dwarfs

Context. Deep learning (DL) techniques are a promising approach among the set of methods used in the ever-challenging determination of stellar parameters in M dwarfs. In this context, transfer learning could play an important role in mitigating uncertainties in the results due to the synthetic gap (i.e. difference in feature distributions between observed and synthetic data). Aims. We propose a feature-based deep transfer learning (DTL) approach based on autoencoders to determine stellar parameters from high-resolution spectra. Using this methodology, we provide new estimations for the effective temperature, surface gravity, metallicity, and projected rotational velocity for 286 M dwarfs observed by the CARMENES survey. Methods. Using autoencoder architectures, we projected synthetic PHOENIX-ACES spectra and observed CARMENES spectra onto a new feature space of lower dimensionality in which the differences between the two domains are reduced. We used this low-dimensional new feature space as input for a convolutional neural network to obtain the stellar parameter determinations. Results. We performed an extensive analysis of our estimated stellar parameters, ranging from 3050 to 4300 K, 4.7 to 5.1 dex, and −0.53 to 0.25 dex for Teff, log 𝑔, and [Fe/H], respectively. Our results are broadly consistent with those of recent studies using CARMENES data, with a systematic deviation in our Teff scale towards hotter values for estimations above 3750 K. Furthermore, our methodology mitigates the deviations in metallicity found in previous DL techniques due to the synthetic gap. Conclusions. We consolidated a DTL-based methodology to determine stellar parameters in M dwarfs from synthetic spectra, with no need for high-quality measurements involved in the knowledge transfer. These results suggest the great potential of DTL to mitigate the differences in feature distributions between the observations and the PHOENIX-ACES spectra.

Wheat Leaf Disease Detection: A Lightweight Approach with Shallow CNN Based Feature Refinement

Improving agricultural productivity is essential due to rapid population growth, making early detection of crop diseases crucial. Although deep learning shows promise in smart agriculture, practical applications for identifying wheat diseases in complex backgrounds are limited. In this paper, we propose CropNet, a hybrid method that utilizes Red, Green, and Blue (RGB) imaging and a transfer learning approach combined with shallow convolutional neural networks (CNN) for further feature refinement. To develop our customized model, we conducted an extensive search for the optimal deep learning architecture. Our approach involves freezing the pre-trained model for feature extraction and adding a custom trainable CNN layer. Unlike traditional transfer learning, which typically uses trainable dense layers, our method integrates a trainable CNN, deepening the architecture. We argue that pre-trained features in transfer learning are better suited for a custom shallow CNN followed by a fully connected layer, rather than being fed directly into fully connected layers. We tested various architectures for pre-trained models including EfficientNetB0 and B2, DenseNet, ResNet50, MobileNetV2, MobileNetV3-Small, and Inceptionv3. Our approach combines the strengths of pre-trained models with the flexibility of custom architecture design, offering efficiency, effective feature extraction, customization options, reduced overfitting, and differential learning rates. It distinguishes itself from classical transfer learning techniques, which typically fine-tune the entire pre-trained network. Our aim is to provide a lightweight model suitable for resource-constrained environments, capable of delivering outstanding results. CropNet achieved 99.80% accuracy in wheat disease detection with reduced training time and computational cost. This efficient performance makes CropNet promising for practical implementation in resource-constrained agricultural settings, benefiting farmers and enhancing production.

Stepwise Transfer Learning for Expert-level Pediatric Brain Tumor MRI Segmentation in a Limited Data Scenario.

Purpose To develop, externally test, and evaluate clinical acceptability of a deep learning pediatric brain tumor segmentation model using stepwise transfer learning. Materials and Methods In this retrospective study, the authors leveraged two T2-weighted MRI datasets (May 2001 through December 2015) from a national brain tumor consortium (n = 184; median age, 7 years [range, 1-23 years]; 94 male patients) and a pediatric cancer center (n = 100; median age, 8 years [range, 1-19 years]; 47 male patients) to develop and evaluate deep learning neural networks for pediatric low-grade glioma segmentation using a stepwise transfer learning approach to maximize performance in a limited data scenario. The best model was externally tested on an independent test set and subjected to randomized blinded evaluation by three clinicians, wherein they assessed clinical acceptability of expert- and artificial intelligence (AI)-generated segmentations via 10-point Likert scales and Turing tests. Results The best AI model used in-domain stepwise transfer learning (median Dice score coefficient, 0.88 [IQR, 0.72-0.91] vs 0.812 [IQR, 0.56-0.89] for baseline model; P = .049). With external testing, the AI model yielded excellent accuracy using reference standards from three clinical experts (median Dice similarity coefficients: expert 1, 0.83 [IQR, 0.75-0.90]; expert 2, 0.81 [IQR, 0.70-0.89]; expert 3, 0.81 [IQR, 0.68-0.88]; mean accuracy, 0.82). For clinical benchmarking (n = 100 scans), experts rated AI-based segmentations higher on average compared with other experts (median Likert score, 9 [IQR, 7-9] vs 7 [IQR 7-9]) and rated more AI segmentations as clinically acceptable (80.2% vs 65.4%). Experts correctly predicted the origin of AI segmentations in an average of 26.0% of cases. Conclusion Stepwise transfer learning enabled expert-level automated pediatric brain tumor autosegmentation and volumetric measurement with a high level of clinical acceptability. Keywords: Stepwise Transfer Learning, Pediatric Brain Tumors, MRI Segmentation, Deep Learning Supplemental material is available for this article. © RSNA, 2024.

Face Warping Deepfake Detection and Localization in a Digital Video using Transfer Learning Approach

Generative AI (GenAI) can generate high-resolution and complex content mimicking the creativity of humans, thereby benefiting industries such as gaming, entertainment, and product design. In recent times, AI-generated fake videos, commonly referred to as deepfakes, have become more commonplace and convincing. An additional deepfake technique, face warping, uses digital processing to noticeably distort shapes on a face. Tracking such warping in images and videos is crucial and preventing its use for destructive purposes. A technique is proposed for detecting and localizing face warped areas in video. The input video is extracted to perform various image pre-processing techniques that refine the video into a format that is more likely to classify the classes efficiently. Transfer learning is employed, and the pre-trained model is adopted to train using Convolutional Neural Network (CNN) with the source videos to identify face warping. Based on the experimental results, it was determined that the proposed model detects and localizes the warped areas of the face satisfactorily with an accuracy of 89.25%.

Unsupervised Transfer Learning Across Different Data Modalities for Bearing's Speed Identification

Recent advancements in transfer learning have revolutionized predictive maintenance, enabling cross-domain generalization for components with varying characteristics and operating under different conditions. While traditional transfer learning approaches require labeled data in both source and target domains, unsupervised transfer learning strives for a more cost-effective alternative for which only labels are available in the source domain. This study investigates adversarial transfer learning between two different sensor modalities: vibration and acoustic. The goal is to enable bearing monitoring using microphones, which are, in general terms, cheaper and easier to deploy than vibration sensors; and without the need to label data in the target domain. The research goal is to identify the operating speed of a bearing testbed. The source domain data correspond to vibration measurements taken from an attached sensor, while the target domain uses a microphone array at distance. Artificial Neural Networks are used as the base architecture. Transferability is assessed with two unsupervised adversarial learning techniques: gradient reversal and deep correlation alignment. Their performance is compared to traditional supervised transfer learning via fine-tuning. Experimental results demonstrate that gradient reversal outperforms deep correlation alignment and is able to achieve results similar to those obtained with supervised transfer learning. These findings highlight the feasibility of speed identification using a microphone array and establish a baseline for future condition monitoring research with such sensors. Conclusion. These findings highlight the feasibility of speed identification using a microphone array and establish a baseline for future condition monitoring research with such sensors.

A Transfer Learning Approach for Identification of Two-Wheeler Brand and Model

A Transfer Learning Approach for Identification of Two-Wheeler Brand and Model