Person: Kumar, Ahlad
Loading...
Name
Ahlad Kumar
Job Title
Email Address
Telephone
Birth Date
Specialization
12 results
Search Results
Now showing 1 - 10 of 12
Publication Metadata only Writer age estimation through handwriting(Springer, 01-05-2023) Huang, Zhiheng; Shivakumara, Palaiahnakote; Kaljahi, Maryam Asadzadeh; Kumar, Ahlad; Pal, Umapada; Lu, Tong; Blumenstein, Michael; Kumar, Ahlad; Kumar, Ahlad; Kumar, Ahlad; Kumar, Ahlad; Kumar, Ahlad; DA-IICT, GandhinagarHandwritten image-based writer age estimation is a challenging task due to the various writing styles of different individuals, use of different scripts, varying alignment, etc. Unlike age estimation using face recognition in biometrics, handwriting-based age classification is reliable and inexpensive because of the plain backgrounds of documents. This paper presents a novel model for deriving the phase spectrum based on the Harmonic Wavelet Transform (HWT) for age classification on handwritten images from 11 to 65�years. This includes 11 classes with an interval of 5�years. In contrast to the Fourier transform, which provides a noisy phase spectrum due to loss of time variations, the proposed HWT-based phase spectrum retains time variations of phase and magnitude. As a result, the proposed HWT-based phase spectrum preserves vital information of changes in handwritten images. In order to extract such information, we propose new phase statistics-based features for age classification based on the understanding that as age changes, writing style also changes. The features and the input images are fed to a VGG-16 model for age classification. The proposed method is tested on our own dataset and three standard datasets, namely, IAM-2, KHATT and that of Basavaraja et al. to demonstrate the effectiveness of the proposed model compared to the existing methods in terms of classification rate. The results of the proposed and existing methods on different datasets show that the proposed method outperforms the existing methods in terms of classification rate.Publication Metadata only Image Denoising Based on Fractional Gradient Vector Flow and Overlapping Group Sparsity as Priors(IEEE, 17-08-2021) Kumar, Ahlad; Ahmad, M Omair; Swamy, M N S; Kumar, Ahlad; Kumar, Ahlad; Kumar, Ahlad; Kumar, Ahlad; Kumar, Ahlad; DA-IICT, GandhinagarIn this paper, a new regularization term in the form of L1-norm based fractional gradient vector flow (LF-GGVF) is presented for the task of image denoising. A fractional order variational method is formulated, which is then utilized for estimating the proposed LF-GGVF. Overlapping group sparsity along with LF-GGVF is used as priors in image denoising optimization framework. The Riemann-Liouville derivative is used for approximating the fractional order derivatives present in the optimization framework. Its role in the framework helps in boosting the denoising performance. The numerical optimization is performed in an alternating manner using the well-known alternating direction method of multipliers (ADMM) and split Bregman techniques. The resulting system of linear equations is then solved using an efficient numerical scheme. A variety of simulated data that includes test images contaminated by additive white Gaussian noise are used for experimental validation. The results of numerical solutions obtained from experimental work demonstrate that the performance of the proposed approach in terms of noise suppression and edge preservation is better when compared with that of several other methods.Publication Metadata only Orthogonal Features Based EEG Signals Denoising Using Fractional and Compressed One-Dimensional CNN Autoencoder(IEEE, 24-08-2022) Nagar, Subham; Kumar, Ahlad; DA-IICT, Gandhinagar; Nagar, Subham (201911004)This paper presents a fractional one-dimensional convolutional neural network (CNN) autoencoder for denoising the Electroencephalogram (EEG) signals which often get contaminated with noise during the recording process, mostly due to muscle artifacts (MA), introduced by the movement of muscles. The existing EEG denoising methods make use of decomposition, thresholding and filtering techniques. In the proposed approach, EEG signals are first transformed to orthogonal domain using Tchebichef moments before feeding to the proposed architecture. A new hyper-parameter (�?�) is introduced which refers to the fractional order with respect to which gradients are calculated during back-propagation. It is observed that by tuning�?�, the quality of the restored signal improves significantly. Motivated by the high usage of portable low energy devices which make use of compressed deep learning architectures, the trainable parameters of the proposed architecture are compressed using randomized singular value decomposition (RSVD) algorithm. The experiments are performed on the standard EEG datasets, namely, Mendeley and Bonn. The study shows that the proposed fractional and compressed architecture performs better than existing state-of-the-art signal denoising methods.Publication Metadata only Tchebichef Transform Domain-Based Deep Learning Architecture for Image Super-Resolution(IEEE, 01-02-2024) Kumar, Ahlad; Singh, Harsh Vardhan; Khare, Vijeta; DA-IICT, Gandhinagar; Singh, Harsh Vardhan (201911022)Recent advances in the area of artificial intelligence and deep learning have motivated researchers to apply this knowledge to solve multipurpose applications in the area of computer vision and image processing. Super-resolution (SR), in the past few years, has produced remarkable results using deep learning methods. The ability of deep learning methods to learn the nonlinear mapping from low-resolution (LR) images to their corresponding high-resolution (HR) images leads to compelling results for SR in diverse areas of research. In this article, we propose a deep learning-based image SR architecture in the Tchebichef transform domain. This is achieved by integrating a transform layer into the proposed architecture through a customized Tchebichef convolutional layer (TCL). The role of TCL is to convert the LR image from the spatial domain to the orthogonal transform domain using Tchebichef basis functions. The inversion of the transform mentioned earlier is achieved using another layer known as the inverse TCL (ITCL), which converts back the LR images from the transform domain to the spatial domain. It has been observed that using the Tchebichef transform domain for the task of SR takes the advantage of high and low- frequency representation of images that makes the task of SR simplified. Furthermore, a transfer learning-based approach is adopted to enhance the quality of images by considering Covid19 medical images as an additional experiment. It is shown that our architecture enhances the quality of X-ray and CT images of COVID-19, providing a better image quality that may help in clinical diagnosis. Experimental results obtained using the proposed Tchebichef transform domain SR (TTDSR) architecture provides competitive results when compared with most of the deep learning methods employed using a fewer number of trainable parameters.Publication Metadata only Orthogonal features-based EEG signal denoising using fractionally compressed autoencoder(Elsevier, 01-11-2021) Nagar, Subham; Kumar, Ahlad; Swamy, M N S; Kumar, Ahlad; Kumar, Ahlad; Kumar, Ahlad; Kumar, Ahlad; Kumar, Ahlad; DA-IICT, Gandhinagar; Nagar, Subham (201911004)A fractional-based compressed auto-encoder architecture has been introduced to solve the problem of denoising�electroencephalogram�(EEG) signals. The architecture makes use of�fractional calculus�to calculate the gradients during the back-propagation process, as a result of which a new hyper-parameter in the form of fractional order��has been introduced which can be tuned to get the best denoising performance. Additionally, to avoid substantial use of memory resources, the model makes use of orthogonal features in the form of Tchebichef moments as input. The orthogonal features have been used in achieving compression at the input stage. Considering the growing use of low energy devices, compression of�neural networks�becomes imperative. Here, the auto-encoder�s weights are compressed using the randomized�singular value decomposition�(RSVD) algorithm during training while evaluation is performed using various compression ratios. The experimental results show that the proposed fractionally compressed architecture provides improved denoising results on the standard datasets when compared with the existing methods.Publication Metadata only Attention-Based Multi-Input Multi-Output Neural Network for Plant Disease Prediction Using Multisensor System(IEEE, 15-12-2022) Saini, Riya; Patle, Kamlesh S; Kumar, Ahlad; Surya, Sandeep G; Palaparthy, Vinay; DA-IICT, Gandhinagar; Saini, Riya (202100101); Patle, Kamlesh S (202121017)Disease detection and prevention in plants are crucial for generating healthy crops and securing the livelihood of farmers. Leaf wetness duration (LWD), ambient temperature, and relative humidity (RH) are essential parameters that lead to the germination of fungal diseases in plants. In this work, an in-house-developed leaf wetness sensor (LWS) is used to capture LWD, and commercial temperature and humidity sensors are used to record the ambient temperature and humidity, respectively. Subsequently, these sensors are interfaced with an in-house-developed Internet of Things (IoT)-enabled electronics and deployed (three sensor nodes) in the field. We have proposed an attention-based multi-input multi-output neural network (A-MIMONN) to predict diseases using self-collected data. The data for training the model are collected from the three sensors nodes, each comprising of temperature, humidity, and LWSs. The designed network is an ensemble of various submodels, trained individually using data from different sensor nodes. The findings of these individually trained networks are then combined to give the final output. The network is designed to achieve better results by employing the attention mechanism, reinforcing the influence of the most important feature on the predicted results. The average accuracy of the model was found to be about 94%. The model displayed a high average precision of 96% and a high average recall value of 97%. The average F1 score of 97% indicated an excellent balance of precision and accuracy.Publication Metadata only Identifying the Source of Water on Plant Using the Leaf Wetness Sensor and via Deep Learning-Based Ensemble Method(IEEE, 01-01-2024) Saini, Riya; Garg, Pooja; Kumar, Naveen Chaudhary; Joshi, Manjunath V; Palaparthy, Vinay; Kumar, Ahlad; DA-IICT, Gandhinagar; Garg, Pooja (202021011)Plant disease detection and management is one of the pivotal areas in the agriculture sector, which needs attention to abate crop loss. The recent trends in machine learning and deep learning have played a significant role in reducing crop loss with the help of early plant disease detection. For plant disease detection prior information on soil moisture, ambient temperature, relative humidity, leaf wetness sensor (LWS), rainfall are crucial parameters. In this work, the objective is to identify the source of leaf wetness on the leaf canopy, which can arise due to irrigation, rainfall, or dew. To identify the source of wetness on the leaf canopy, either rainfall or humidity/mist sensors are used, which substantially increases the cost of the system. For this purpose, we have used the LWS, which is deployed in the field and various patterns for the irrigation, rainfall, or dew has been analyzed by using the in-house developed the Internet of Things (IoT)-enabled sensor system. The data collected from the field is used as a learning dataset for the proposed ensemble neural network (NN) developed to identify the source of leaf wetness. Short-time Fourier transform (STFT) has been employed to enhance data representation by transforming numerical data from the LWS into informative images. The provided ensemble model incorporates convolutional NN (CNN) and multilayer perceptron (MLP), which process image and numerical data (ambient temperature, relative humidity, leaf wetness duration, and maximum magnitude of frequency of images) as input. Their outputs combined in an artificial neural network (ANN) sub-model for precise leaf wetness event detection (dew, rainfall, or irrigation). The proposed model achieved an accuracy of 96.13% with average precision, recall, and F1 score for the leaf wetness events is about 84%, 85%, and 83%, respectively.Publication Metadata only IoT Enabled, Leaf Wetness Sensor on the Flexible Substrates for In-Situ Plant Disease Management(IEEE, 16-06-2021) Patle, Kamlesh S; Saini, Riya; Kumar, Ahlad; Surya, Sandeep G; Palaparthy, Vinay; Salama, Khaled N; DA-IICT, Gandhinagar; Patle, Kamlesh S (202121017)Early plant disease detection and providing the control measures have become highly desirable to improve crop yield. Leaf wetness duration (LWD) is one of the essential parameters related to the development of fungal disease on the leaf canopy. To measured LWD, the leaf wetness sensor (LWS) is widely used. Commercially available LWS are made on printed circuit board (PCB) technology, which has an operational issue during field deployment such as weight of the sensor, contact resistance between the sensor and the leaves, form factor and most importantly, affordability. To mitigate the issues associated with the commercially available LWS, in this work, we have fabricated the in-house IoT-enabled and affordable electronic leaf wetness sensor on the flexible substrates, which is used for integrated plant disease management. Fabricated LWS comprises the interdigitated electrodes (IDEs) on the polyimide flexible substrate. The lab measurement results indicate that fabricated LWS on the flexible substrates offers a response of about 36000% when LWS is exposed to water w.r.t air. The observed response time of the fabricated LWS is about 10 seconds and hysteresis of about � 4 %. Further, sensor capacitance changes only by 6% over a temperature range from 20 �C to 65 �C. Furthermore, three fabricated sensors LWS and in-house developed internet of things (IoT) enabled systems are deployed on the Ocimum tenuiflorum (Tulsi) medical plant. Field measurement indicates that measured LWD using the fabricated flexible LWS and commercially available LWS (Phytos 31:LWS-L12), METER Group, Inc. USA) shows the absolute difference of 30 minutes.Publication Metadata only Field Evaluation of Smart Sensor System for Plant Disease Prediction using LSTM Network(IEEE, 15-02-2022) Patle, Kamlesh S; Saini, Riya; Kumar, Ahlad; Palaparthy, Vinay; DA-IICT, Gandhinagar; Patle, Kamlesh S (202121017); Saini, Riya (202100101)Leaf wetness duration (LWD), soil moisture, soil temperature, ambient temperature, and relative humidity information are the essential factors that leads to germination of plant disease. In this work, an internet of things (IoT) enabled leaf wetness sensor (LWS) and soil moisture sensor (SMS) is developed. Subsequently, commercial soil temperature (ST), relative humidity (RH) and ambient temperature (AT) are used for plant disease prediction. The developed LWS offers a response of about 250% when exposed to air and water and response time of about 20 seconds and attributes a hysteresis of about �3 %. Acrylic protective lacquer (APL) coating of about 25-�75�?m�thin is deposited on LWS and it is observed that the sensor capacitance changes only by 2% when temperature varies from 20 �C to 65 �C. Likewise, fabricated SMS offers a response of 10 kHz (�?F�) with only a 2% change in frequency when temperature varies from 20 �C to 65 �C and works with an accuracy of �3%. Further, aforementioned sensors along with in-house developed IoT-enabled system has been deployed under field conditions for about four months. In this work, we considered Powdery mildew (D1), Anthracnose (D2), and Root rot (D3) disease on the Mango plant. Further, we have implemented the Long Short Term Memory (LSTM) network which performs better compared to the existing methods discussed on plant disease management. The proposed network achieves an accuracy of 96%, precision-recall and F1 score of 97%, 98%, and 99%, respectively.Publication Metadata only Soil Sensors-Based Prediction System for Plant Diseases Using Exploratory Data Analysis and Machine Learning(IEEE, 15-08-2021) Kumar, Manish; Kumar, Ahlad; Palaparthy, Vinay; DA-IICT, GandhinagarPlant diseases cause losses to agricultural production and hence, the economy. This necessitates a need to develop prediction models for the plant disease detection and assessment. Fungal infection, the most dominant disease, can be controlled by taking appropriate measures if detected at an early stage. The article aims to develop an expert system for the prediction of various fungal diseases (powdery mildew, anthracnose, rust, and root rot/leaf blight). A multi-layered perceptron model is used for the classification of the diseases which not only detects the plant diseases effectively but can also increase the production drastically. The proposed technique incorporates three significant steps of dataset pre-processing, exploratory data analysis, and detection module. Firstly, the real-time data is captured by the soil sensors system installed at agriculture field at Sardarkrushinagar Dantiwada Agricultural University, Gujarat, India, along with the satellite data for other micro-meteorological factors. Next, an extensive exploratory data analysis has been performed to get insights into the collected data. Finally, the proposed machine learning model has been employed to predict plant diseases. The experimental results indicate that the model outperforms several existing methods in terms of accuracy. Average accuracy in predicting each disease has been found more than 98%. This work also proves the feasibility of using this technique for faster plant disease detection at an affordable cost.