Hybrid Dataset Research Articles

A new empirical formulation for the out-of-plane resistance of masonry infills in reinforced concrete frames

The paper presents a new empirical expression estimating the out-of-plane (OOP) resistance of infilled reinforced concrete frames subject to horizontal forces. The new model is calibrated through an optimization process based on a hybrid dataset, including experimental data from real tests and from numerical simulations obtained from a refined FE micro-model realized in Abaqus®. The new expression considers the effect of vertical loads and also introduces a conversion factor to uniformize point-load and uniform load out-of-plane tests. The final expression is also flexible with respect to available data on the infill material properties, as it is specialized in two versions, one providing the conventional compressive strength of the units, the other providing the conventional compressive strength of masonry as possible inputs. Results presented in the paper will show a noticeable accuracy of the proposed model in estimating the ultimate out-of-plane load of a masonry infill wall, with respect to available models. The formula provides the reference undamaged OOP resistance and can be easily combined with available strength-reduction functions to consider the effect of prior in-plane loading.

Understanding urban wealth perception using a hybrid dataset and ranking‐scoring framework

AbstractResearch on urban perception using street view images has developed rapidly. Global urban perception datasets commonly used in local urban perception research may be inaccurate because of the differences in the characteristics across cities. Building a new local urban perception dataset requires a practical and efficient solution to construct high‐availability datasets and a corresponding urban perception framework. In this study, we developed a simple and effective ranking‐scoring framework containing two stages. First, we constructed a hybrid dataset in a 1:1 ratio, based on the global and local datasets. Next, we built a ranking‐scoring network that uses deep learning models to obtain the features of images and a learning‐to‐rank approach to determine perceptual scores. To validate the effectiveness of our method, we examined the perception of urban wealth in Wuhan, China. Our framework showed high accuracy when the hybrid dataset was used. Furthermore, we analyzed the differences between the spatial distribution of the perception results of the hybrid model and the global model in the study area. In addition, we analyzed driving factors for wealth perception, such as visual elements and road types. The results verified that the hybrid dataset is more suitable for local urban perception. Our framework could assist researchers in conducting fast and accurate urban perception of local cities, using a hybrid dataset and ranking‐scoring, thus enabling urban planners to provide more consideration to users' perceptions during the planning process without dwelling upon additional burdens.

Predicting quorum sensing peptides using stacked generalization ensemble with gradient boosting based feature selection.

Bacteria exist in natural environments for most of their life as complex, heterogeneous, and multicellular aggregates. Under these circumstances, critical cell functions are controlled by several signaling molecules known as quorum sensing (QS) molecules. In Gram-positive bacteria, peptides are deployed as QS molecules. The development of antibodies against such QS molecules has been identified as a promising therapeutic intervention for bacterial control. Hence, the identification of QS peptides has received considerable attention. Availability of a fast and reliable predictive model to effectively identify QS peptides can help the existing high throughput experiments. In this study, a stacked generalization ensemble model with Gradient Boosting Machine (GBM)-based feature selection, namely EnsembleQS was developed to predict QS peptides with high accuracy. On selected GBM features (791D), the EnsembleQS outperformed finely tuned baseline classifiers and demonstrated robust performance, indicating the superiority of the model. The accuracy of EnsembleQS is 4% higher than those resulting from ensemble model on hybrid dataset. When evaluating an independent data set of 40 QS peptides, the EnsembleQS model showed an accuracy of 93.4% with Matthew's Correlation Coefficient (MCC) and area under the ROC curve (AUC) values of 0.91 and 0.951, respectively. These results suggest that EnsembleQS will be a useful computational framework for predicting QS peptides and will efficiently support proteomics research. The source code and all datasets used in this study are publicly available at https://github.com/proteinexplorers/EnsembleQS .

DOPE++: 6D pose estimation algorithm for weakly textured objects based on deep neural networks.

This paper focuses on 6D pose estimation for weakly textured targets from RGB-D images. A 6D pose estimation algorithm (DOPE++) based on a deep neural network for weakly textured objects is proposed to solve the poor real-time pose estimation and low recognition efficiency in the robot grasping process of parts with weak texture. More specifically, we first introduce the depthwise separable convolution operation to lighten the original deep object pose estimation (DOPE) network structure to improve the network operation speed. Second, an attention mechanism is introduced to improve network accuracy. In response to the low recognition efficiency of the original DOPE network for parts with occlusion relationships and the false recognition problem in recognizing parts with scales that are too large or too small, a random mask local processing method and a multiscale fusion pose estimation module are proposed. The results show that our proposed DOPE++ network improves the real-time performance of 6D pose estimation and enhances the recognition of parts at different scales without loss of accuracy. To address the problem of a single background representation of the part pose estimation dataset, a virtual dataset is constructed for data expansion to form a hybrid dataset.

Web Application Firewall Using Machine Learning and Features Engineering

Web application security has become a major requirement for any business, especially with the wide web attacks spreading despite the defensive measures and the continuous development of software frameworks and servers. In this study, we present a proposed model for a web application firewall that used machine learning and features engineering to detect common web attacks. Our proposed model analyses incoming requests to the webserver, parses these requests to extract four features that describe completely HTTP request parts (URL, payload, and headers), and classifies whether a request is normal or an anomaly. We took into consideration the limitation of previous works that use URL and payload only in classification and provided five features that describe and summarize all parts of the HTTP request using features engineering and previous experience in the field of the software security domain. Extracted features are length of request, percentage of characters allowed, percentage of special characters, and attack weight. These features were calculated for four different datasets CSIC 2010, HTTPParams 2015, Hybrid dataset (CSIC 2010 and HTTPParams), and real logs for the compromised web server. We evaluated our proposed model by using these updated datasets with four classification algorithms (Naive Bayes, logistic regression, decision tree, and support vector machine) with two methods (train test split and cross-validation) to negate the probability of overfitting and ensure that features are effective. Features values for a normal request are usually short request length, large allowed character ratio, small special character ratio, and zero attack weight or close to zero. Features values for anomaly requests are large request length, small allowed character percentage, large special character percentage, and very large numerically attack weight. Our proposed model achieved a classification accuracy of 99.6% with datasets used in research studies in this field and 98.8% with datasets of real web servers.

Machine learning-based risk prediction model for cardiovascular disease using a hybrid dataset

CVD (cardiovascular disease) is one of the most common causes of death in the world today. CVD prediction allows health professionals to make an informed decision about their patients’ health. Data mining is the process of transforming large amounts of medical data in its raw form into actionable insights that can be used to make intelligent forecasts and decisions. Machine learning (ML) based prediction models provide a better solution to help patients’ health diagnoses in the health care industry. The objective of this research is to create a hybrid dataset to aid in the development of a best CVD risk prediction model. The Hungarian, the Switzerland, the Cleveland, and the Long Beach datasets are the most commonly used datasets in heart disease (HD) prediction. These datasets have a maximum of 303 instances with missing values in their features, and the presence of missing values reduces the accuracy of the prediction model. So, in this article, we created the ”Sathvi” dataset by combining these datasets, and it has 531 instances with 12 attributes with no missing data. The Pearson’s correlation method was used to eliminate redundant features during the feature selection process. The Naive Bayes (NB), XGBoost, k-nearest neighbour (k-NN), multilayer perceptron (MLP), support vector machine (SVM), and CatBoost ML classifiers have been applied for prediction. The CatBoost ML classifier was validated with 10-fold cross validation, and the best accuracy ranged from 88.67% to 98.11%, with a mean of 94.34%.

Road and Railway Smart Mobility: A High-Definition Ground Truth Hybrid Dataset.

A robust visual understanding of complex urban environments using passive optical sensors is an onerous and essential task for autonomous navigation. The problem is heavily characterized by the quality of the available dataset and the number of instances it includes. Regardless of the benchmark results of perception algorithms, a model would only be reliable and capable of enhanced decision making if the dataset covers the exact domain of the end-use case. For this purpose, in order to improve the level of instances in datasets used for the training and validation of Autonomous Vehicles (AV), Advanced Driver Assistance Systems (ADAS), and autonomous driving, and to reduce the void due to the no-existence of any datasets in the context of railway smart mobility, we introduce our multimodal hybrid dataset called ESRORAD. ESRORAD is comprised of 34 videos, 2.7 k virtual images, and 100 k real images for both road and railway scenes collected in two Normandy towns, Rouen and Le Havre. All the images are annotated with 3D bounding boxes showing at least three different classes of persons, cars, and bicycles. Crucially, our dataset is the first of its kind with uncompromised efforts on being the best in terms of large volume, abundance in annotation, and diversity in scenes. Our escorting study provides an in-depth analysis of the dataset’s characteristics as well as a performance evaluation with various state-of-the-art models trained under other popular datasets, namely, KITTI and NUScenes. Some examples of image annotations and the prediction results of our 3D object detection lightweight algorithms are available in ESRORAD dataset. Finally, the dataset is available online. This repository consists of 52 datasets with their respective annotations performed.

Reverse Image Search Using Deep Unsupervised Generative Learning and Deep Convolutional Neural Network

Reverse image search has been a vital and emerging research area of information retrieval. One of the primary research foci of information retrieval is to increase the space and computational efficiency by converting a large image database into an efficiently computed feature database. This paper proposes a novel deep learning-based methodology, which captures channel-wise, low-level details of each image. In the first phase, sparse auto-encoder (SAE), a deep generative model, is applied to RGB channels of each image for unsupervised representational learning. In the second phase, transfer learning is utilized by using VGG-16, a variant of deep convolutional neural network (CNN). The output of SAE combined with the original RGB channel is forwarded to VGG-16, thereby producing a more effective feature database by the ensemble/collaboration of two effective models. The proposed method provides an information rich feature space that is a reduced dimensionality representation of the image database. Experiments are performed on a hybrid dataset that is developed by combining three standard publicly available datasets. The proposed approach has a retrieval accuracy (precision) of 98.46%, without using the metadata of images, by using a cosine similarity measure between the query image and the image database. Additionally, to further validate the proposed methodology’s effectiveness, image quality has been degraded by adding 5% noise (Speckle, Gaussian, and Salt pepper noise types) in the hybrid dataset. Retrieval accuracy has generally been found to be 97% for different variants of noise

DMS-SK/BLSTM-CTC Hybrid Network for Gesture/Speech Fusion and Its Application in Lunar Robot–Astronauts Interaction

In the future manned lunar exploration mission, astronauts would work with the lunar robots, which has a high requirement for human–robot interaction (HRI). As the accuracy of gesture recognition interaction does not fulfill the requirement for human–robot joint exploration missions, we propose the DMS-SK/BLSTM-CTC hybrid network to improve the performance of HRI. For gesture recognition, considering VGG-SK has low accuracy and complex architecture, we delete the fourth convolution module, optimize the last global pooling layer, introduce dilated convolution block and multiscale convolution block in VGG-SK, and get the DMS-SK-based gesture recognition sub-network. Compared with the traditional recognition methods, the accuracy and performance of DMS-SK improve. For speech recognition, considering that Bidirectional long–short-term memory unit (BLSTM) has the advantages of processing temporal information, and the Connectionist Temporal Classification (CTC) algorithm can simplify speech data preprocessing, we use BLSTM based on CTC as the speech recognition sub-network. Finally, we combine DMS-SK with BLSTM-CTC, and propose the DMS-SK/BLSTM-CTC hybrid network as the gesture/speech hybrid network. In addition, we use 10 gestures in the American Sign Language (ASL) dataset and 10 speech commands to construct the gesture/speech hybrid dataset. Experimental results show that compared with the pure gesture or pure speech networks, the recognition accuracy of the gesture-speech hybrid network improves by 2% and 12%, respectively, its accuracy reaches 97.38%, which fulfills the requirement of astronauts for HRI.

Indian Monument Recognition Using Deep Learning

Monuments are physical structures built or created dedicated to a person or event. Their importance to the region they belong to makes their documentation and preservation important. Monuments being 3D objects, pose is an issue in terms of the difference in perception of the monument images. There can be an immense variation because of the viewpoint from which the image is being taken. To overcome this issue, Data Augmentation is used in this study. Random Flip, Random Rotation, Random Translation and Random Zoom, Random Contrast, Random Hue, Random Brightness, and Random Saturation operations under Data Augmentation are applied on the dataset. This research uses Deep Learning architectures on a hybrid dataset of “Indian Monument Recognition Dataset” and “Qutub Complex Monuments' Images Dataset.” InceptionV3, MobileNet, ResNet50, AlexNet, and VGG16 recognize the monuments with an accuracy of 97.79%, 93.73%, 86.47%, 68.88%, and 61.33% respectively.

Timing shift-based bi-residual network model for the detection of electricity stealing

With the increasing number of electricity stealing users, the interests of countries are jeopardized and it brings economic burden to the government. However, due to the small-scale stealing and its random time coherence, it is difficult to find electricity stealing users. To solve this issue, we first generate the hybrid dataset composed of real electricity data and specific electricity stealing data. Then, we put forward the timing shift-based bi-residual network (TS-BiResNet) model. It learns the features of electricity consumption data on two aspects, i.e., shallow features and deep features, and meanwhile takes time factor into consideration. The simulation results show that TS-BiResNet model can detect electricity stealing behaviors that are small scaled and randomly coherent with time. Besides, its detection accuracy is superior to the benchmark schemes, i.e., long short-term memory (LSTM), gated recurrent unit (GRU), combined convolutional neural network and LSTM (CNN-LSTM) and Bi-ResNet.

Generalization of solar power yield modeling using knowledge transfer

Recently, the energy issue is an important factor all over the world due to the fact of sustainability perspective. This paper has proposed a new Transfer Learning (TL) based PV energy yield model, which generalizes the knowledge transfer process from the source domain to the target domain differ on time and location. The solar power yield forecasting highly depends on meteorological data, which has high variability influenced by both time and target place/location. Therefore, a solar energy forecasting model from a source may not perform well to a target domain due to differences in the data distribution as evident by the experiment result presented in this paper. The extended historical data-driven solar power yield models may perform better to a certain extent with the complex nonlinear meteorological data relationships. But due to rapid climate change issue, the knowledge learned from historical data may not apply to the near future. Furthermore, it is impractical to assume the availability of a large dataset and ignore the probability of negative knowledge transfer tendency of traditional regression models. To address this gap, this paper has proposed Transfer Support Vector Regression (Tr-SVR) that uses a novel weighting model based on patterns of a hybrid dataset, both source and target data for training. The proposed Tr-SVR only transfer the knowledge fits the target domain to reduce negative transfer and dependency on long historical data. The experimental results show that when the distribution of training and test data differs, the traditional machine learning method represented by ANN may face serious generalization challenges. In this scenario, the proposed Tr-SVR model has clear advantages in terms of error metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Maximum error, Mean Absolute Percentage Error (MAPE), and Sum of Squares of the Error (SSE).

A novel principal component-based virtual sensor approach for efficient classification of gases/odors

Abstract High-performance detection and estimation of gases/odors are challenging, especially in real-time gas sensing applications. Recently, efficient electronic noses (e-noses) are being developed using convolutional neural networks (CNNs). Further, CNNs perform better when they operate on a minimal size of vector response. In this paper, dimensions of the operational vectors have been augmented by using virtual sensor responses. These virtual responses are obtained from the principal components of the physical sensor responses. Accordingly, two sets of data are upscaled as a one-dimensional one. Another level of upscaling is further obtained by using the mirror mosaicking technique. Hence, with our proposed novel approach, the final vector size for CNN operations achieves a new dimension. With this upscaled hybrid dataset, consisting of physical and virtual sensor responses, a simpler CNN has achieved 100 percent correct classification in two different experimental settings. To the best of authors information, it is for the first time that an e-nose has been designed using a principal component-based hybrid, upscaled dataset and achieves 100 percent correct classification of the considered gases/odors.

Identifying Country-Level Risk Factors for the Spread of COVID-19 in Europe Using Machine Learning.

Coronavirus disease 2019 (COVID-19) has resulted in approximately 5 million deaths around the world with unprecedented consequences in people’s daily routines and in the global economy. Despite vast increases in time and money spent on COVID-19-related research, there is still limited information about the factors at the country level that affected COVID-19 transmission and fatality in EU. The paper focuses on the identification of these risk factors using a machine learning (ML) predictive pipeline and an associated explainability analysis. To achieve this, a hybrid dataset was created employing publicly available sources comprising heterogeneous parameters from the majority of EU countries, e.g., mobility measures, policy responses, vaccinations, and demographics/generic country-level parameters. Data pre-processing and data exploration techniques were initially applied to normalize the available data and decrease the feature dimensionality of the data problem considered. Then, a linear ε-Support Vector Machine (ε-SVM) model was employed to implement the regression task of predicting the number of deaths for each one of the three first pandemic waves (with mean square error of 0.027 for wave 1 and less than 0.02 for waves 2 and 3). Post hoc explainability analysis was finally applied to uncover the rationale behind the decision-making mechanisms of the ML pipeline and thus enhance our understanding with respect to the contribution of the selected country-level parameters to the prediction of COVID-19 deaths in EU.

MUSAK: A Multi-Scale Space Kinematic Method for Drone Detection

Accurate and robust drone detection is an important and challenging task. However, on this issue, previous research, whether based on appearance or motion features, has not yet provided a satisfactory solution, especially under a complex background. To this end, the present work proposes a motion-based method termed the Multi-Scale Space Kinematic detection method (MUSAK). It fully leverages the motion patterns by extracting 3D, pseudo 3D and 2D kinematic parameters at three scale spaces according to the keypoints quality and builds three Gated Recurrent Unit (GRU)-based detection branches for drone recognition. The MUSAK method is evaluated on a hybrid dataset named multiscale UAV dataset (MUD), consisting of public datasets and self-collected data with motion labels. The experimental results show that MUSAK improves the performance by a large margin, a 95% increase in average precision (AP), compared with the previous state-of-the-art (SOTA) motion-based methods, and the hybrid MUSAK method, which integrates with the appearance-based method Faster Region-based Convolutional Neural Network (Faster R-CNN), achieves a new SOTA performance on AP metrics (AP, APM, and APS).

A Hybrid Dataset of Historical Cool-Season Lake Effects From the Eastern Great Lakes of North America

The moistening of cold air passing over the Great Lakes of North America has a profound impact on the cool season climate of regions downwind, from relatively benign air mass modification to highly-impactful snowfall events. The importance of lake effects has led to the development of varying techniques for systematically identifying lake-effect days. The results of two such methods are merged here to yield a more thorough record of lake-effect days for the eastern Great Lakes. Comparative analysis of the data sets illustrates the different objectives of the two methodologies, where one identifies days with a synoptic setup conducive to lake-effect snowfall, and the other identifies days with lake-effect modification of the overlying air mass. A smaller population of “absolute” lake-effect days are those identified by both methods, while a larger population of “hybrid” lake-effect days are absolute days plus those identified by one method but not the other. For a 51-year study period ending with the 2014–15 cool season, the absolute data set yields a mean of about 15 lake-effect days per year, or 8% of the November through April season, while the hybrid data set yields a mean of 56 lake-effect days per year, or 31% of the season. The frequencies of absolute, air mass modification-defined, and hybrid lake-effect days decreased through the study period, with days within the hybrid data set declining at a statistically significant rate of 2.8 days per decade, although most obviously from the late 1970s through the early 2000s. The result is a general drying of the cool-season lake-effect hydroclimate. The merged data set offers a more thorough historical record of days available for atmospheric and hydroclimatic study of the lake-effect phenomenon within the eastern Great Lakes region.

Spatiotemporal Deep-Learning-Based Algal Bloom Prediction for Lake Okeechobee Using Multisource Data Fusion

This study focuses on predicting harmful algal bloom (HAB) events in Lake Okeechobee, a shallow lake in Florida. A spatio-temporal deep learning model is employed to predict the levels of cyanobacteria <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Microcystis aeruginosa</i> (M. aeruginosa) present in the lake for a single-day and a 14-day prediction horizon. Datasets collected from remote sensing (i.e., satellite images from Jan. 2018 to Dec. 2020) and from a physics-based simulation model (i.e., daily simulation from Jan. 2018 to Dec. 2020) are available. Due to the low quality of remote sensing data caused by various environmental and technical issues, the two available datasets are fused together to create a multi-source hybrid dataset for deep learning model training. A convolutional long-short term memory (ConvLSTM) deep neural model is trained on the datasets, and the results of the predictions are compared to the true Cyanobacterial Index (CI) for that time period. Findings include 1) the deep learning model, ConvLSTM, shows promising performance for short- and mid-term HAB forecasting; and 2) the hybrid dataset that fuses remote sensing with physics-based modeling (a.k.a. modeling based on fundamental physical and biogeochemical principles) speeds up the model learning and improves its performance significantly. The proposed methodologies are reliable, and cost-effective, and could be used to forecast algal bloom occurrences in shallow lakes with limited sparse observations.

Improving Indoor Localization Using Mobile UWB Sensor and Deep Neural Networks

Accurate localization in indoor environments with ultra-wideband (UWB) technology has long attracted much attention. However, due to the presence of multipath components or non-line of sight (NLOS) propagation of the radio signals, it has been converted to a critical challenge. Existing solutions use many fixed anchors in the indoor environment. Particularly, large areas require many anchor points and in the case of unexpected events that lead to the destruction of existing infrastructures, the fixed anchor points cannot be used. In this paper, a novel localization framework based on the transmitting signal from a mobile UWB sensor on the outside of the building and its received signal regarding the modified Saleh Valenzuela (SV) channel model is presented. After preprocessing the received signals, two new procedures to reduce the ranging error caused by multipath components are proposed. In the first procedure, two machine learning algorithms including multi-layer perceptron (MLP) and support vector machine (SVM) using the extracted features from the received UWB signal time and power vectors are implemented. Moreover, in the second procedure, two deep learning algorithms including MLP and convolutional neural networks (CNNs) using the received UWB signal time and power vectors are implemented to improve the performance of the indoor localization system. The simulation results show that the architecture designed for the convolutional neural network based on the hybrid dataset (the combination of the dataset related to received UWB signal time and power vectors) provides a mean absolute error (MAE) of about 3 cm.

A Single Stream Modified MobileNet V2 and Whale Controlled Entropy Based Optimization Framework for Citrus Fruit Diseases Recognition

Fruit disease recognition is quickly becoming a hot topic in the field of computer vision. The presence of plant diseases not only reduces fruit production but also causes a significant loss to the national economy. Citrus fruits help to strengthen the immune system, allowing it to fight off diseases such as COVID-19. Manual inspection of fruit diseases with the naked eye takes time and is difficult; therefore, a computer based method is always required for accurate recognition of plant diseases. Several deep learning techniques for recognizing citrus fruit diseases have been introduced in the literature. Existing techniques had several issues, including redundant features, convolutional neural network (CNN) model selection, low contrast images, and long computational times. In this paper, a single stream convolutional neural network architecture is proposed for recognizing citrus fruit diseases. In the first step, data augmentation is performed using four contrast enhancement operations: shadow removal, adjusting pixel intensity, improving brightness, and improving local contrast. The MobileNet-V2 CNN model is selected and fine-tuned in the second step. Using the transfer learning process, the fine-tuned model is trained on the augmented citrus dataset. The newly trained model is used for deep feature extraction; however, analysis shows that the extracted deep features contain little redundant information. As a result, an improved Whale Optimization Algorithm (IWOA) is used in the third step. The best features are then classified using machine learning classifiers in the final step. The augmented citrus fruits, leaves, and hybrid dataset were used in the experimental process and achieved an accuracy of 99.4, 99.5, and 99.7%. When compared to existing techniques, the proposed architecture outperformed them in terms of accuracy and time.

On-device Resilient Android Malware Detection using Incremental Learning

In most cases, the Android runs in a battery-driven, low-power connected, embedded real-time environment. New malware attacks are reported every day targeting Android embedded devices. This is because the existing Android malware detection engines are not resilient to zero day attacks. The machine learning based malware detection engines must be re-trained with the newly reported malware and benign samples to manage the zero-day attacks. The re-training is very expensive in terms of CPU and memory. Re-training at embedded devices is very challenging because of its hardware, memory, and energy constraints. In these scenarios, a lightweight, resilient and incremental learning based dynamic Android malware detection mechanism is becoming mandatory. Very little evidence is observed in the related literature; most of the existing incremental learning approaches have not demonstrated their experiments in a real device environment and have not worked in bench-marked real device dynamic analysis datasets. In this work, we select the latest KronoDroid Android dynamic hybrid dataset and propose efficient incremental learning based Android malware detection mechanism using the linear incremental algorithm. We have incrementally trained our mechanism in the real device Android environment and proved the consistent performance keeping the resilient aspect. We have evaluated our mechanism and found that we could achieve the highest incremental accuracy of 97.94%, the lowest training time of 0.5 seconds, and lowest detection time of 0.04 seconds with the smallest model size of 456 bytes when experimented within an on-device Android environment. To the best of our knowledge, this is the first work that proves the incremental learning and resilience scenario using a bench-marked Android real-device dynamic dataset keeping the lowest computational complexity and consistent performance.