Rashi Agarwal
@kanpuruniversity.org
Head. Department of Information Technology
UIET, CSJMU, Kanpur
RESEARCH INTERESTS
Computer Vision, Machine Learning, Image processing, Deep Learning
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Vishal Awasthi, Namita Awasthi, Hemant Kumar, Shubhendra Singh, Prabal Pratap Singh, Poonam Dixit, and Rashi Agarwal
Elsevier BV
Hemant Kumar, Abhishek Dwivedi, Abhishek Kumar Mishra, Arvind Kumar Shukla, Brajesh Kumar Sharma, Rashi Agarwal, and Sunil Kumar
Elsevier BV
Hemant Kumar and Rashi Agarwal
Springer Nature Singapore
S. Hariharan and Rashi Agarwal
Springer Nature Singapore
Rashi Agarwal
IEEE
Diabetes Retinopathy is the primary reason for avoidable vision loss, especially among people of working age. Diabetes is expected to affect 600 million people worldwide by 2040, with one-third developing Diabetic Retinopathy [1] Recent studies have highlighted the need for more cost-effective and efficient methods of diagnosing, managing, and treating retinal diseases in clinical eye care settings. Before the development of a computer-aided diagnosis tool, the difficulty of achieving an accurate early diagnosis at an affordable cost must be considered. It may be possible to expand diabetic retinopathy screening efforts and allow physicians to use their time more efficiently by utilizing computer-aided disease detection in retinal image processing. Biomedical engineers and computer scientists now have a unique opportunity to meet the demands of clinical practice, thanks to advances in edge computing and machine learning techniques. We have used Enhanced U-Net to segment typical pathological features of DR which include hard exudates (EX), soft exudates (SE), microaneurysms (MA), and haemorrhages (HE) in the Indian Diabetic Retinopathy Image Dataset (IDRiD) with better accuracy than other techniques used in this domain. The Enhanced U-Net is able to segment all the features at an early stage, hence, this method can be used to aid diagnosis of DR at an early stage, thereby preventing vision loss at later stages.
Prateek Srivastava, Nidhi Srivastava, Rashi Agarwal, and Pawan Singh
IEEE
Agile software development has a good track record, but only when the project size is limited. Requirement engineering (RE) is critical because “Requirements” play a critical role at each stage of the lifecycle of software. Agile supports changing customer requirements while adding value to the business, it also comes with its own set of obstacles, many of which are connected to requirement constraints. This study seeks to uncover the issues that agile project requirements engineers confront. Many research studies on requirement issues have been undertaken, all of which are compromised in some way, offer no suggestions for improving the agile development process, and fail to reveal massive agile development hurdles. This paper addresses all the issues raised above and suggests possible solutions from the standpoint of requirement engineering and results can be extremely beneficial to the software business in terms of improving development processes, as well as individuals who would like to continue working in this area.
Rashi Agarwal and S. Hariharan
Springer Nature Singapore
Rashi Agarwal, Silky Goel, and Rahul Nijhawan
IEEE
The land cover is the apparent (bio)physical cover, and land use alludes to how the actual land type is being utilized. This research is fundamental to survey the degree to which social, monetary, and natural factors influence urbanization. This will likewise assist with urban planning. As laborious process of handcrafted feature extraction has not helped obtain high accuracies, this paper proposes use of Deep Learning approach that explores different Image Recognition Models using various ML classifiers on remote sensing images classifying the images from large Landsat satellite dataset into 9 different classes. It was observed that the highest accuracy of 97.4% was achieved by the Logistic Regression algorithm coupled with Inceptionv3 model. The proposed model shows the capability of increasing the accuracy of existing state-of-art-algorithms low resolution land classification maps. Thus, the improved results will contribute to better land maps helping with the growing demand of LULC information concerning climate change and sustainable development.
Sunil Kumar, Harish Kumar, Rashi Agarwal, and V. K. Pathak
Springer Nature Singapore
Prateek Srivastava, Nidhi Srivastava, Rashi Agarwal, and Pawan Singh
Springer Singapore
Anand Handa, Rashi Agarwal, and Narendra Kohli
IGI Global
Due to the highly variant face geometry and appearances, Facial Expression Recognition (FER) is still a challenging problem. CNN can characterize 2-D signals. Therefore, for emotion recognition in a video, the authors propose a feature selection model in AlexNet architecture to extract and filter facial features automatically. Similarly, for emotion recognition in audio, the authors use a deep LSTM-RNN. Finally, they propose a probabilistic model for the fusion of audio and visual models using facial features and speech of a subject. The model combines all the extracted features and use them to train the linear SVM (Support Vector Machine) classifiers. The proposed model outperforms the other existing models and achieves state-of-the-art performance for audio, visual and fusion models. The model classifies the seven known facial expressions, namely anger, happy, surprise, fear, disgust, sad, and neutral on the eNTERFACE’05 dataset with an overall accuracy of 76.61%.
Rashi Agarwal, S Hariharan, M. Nagabhushana Rao, and Abhishek Agarwal
IEEE
Food security is a pertinent global challenge that has plagued the nations over time immemorial. Maize is one of the world's most significant consumption crop, based on production volume. Maize supply can vary according to the cultivation area, climatic condition, and disease. Modern precision weed management relies on site-specific management tactics to maximize resource use efficiency and yield, while reducing unintended environmental impacts caused by herbicides. Recent advancements in Unmanned Aircraft Systems (UAS)-based tools and geospatial information technology have created enormous applications for efficient and economical assessment of weed infestations as well as site-specific weed management. This paper explores the possibility of extracting features from color spaces and combining them with vegetation indices (NDVI, VARI, and TGI) to be clustered using K-means classifier to identify the weed population from a multispectral imagery. The results give a clear indication that the NDVI performance is better than VARI; It also shows that TGI is not acceptable for the classification.
Rashi Agarwal, Rahul Nijhawan, Siddharth Gupta, and Silky Goel
IEEE
Covid-19 has quickly emerged as a global threat, tipping the world into a new phase. The delay in medical care because of the quickly rising Covid-19 cases makes it necessary to overcome the manual and time taking technique such as RTPCR. This paper implements different pre-trained CNN feature extraction models using various Machine Learning (ML) classifiers on chest CT scans to analyze Covid-19 infected patients. It may be observed from the obtained results that accuracy of 96.4% was obtained using the VGG16 model and neural network classifier. The implementation of pre-trained models and classifiers reduce the time taken for manual detection of disease and helps doctors to prevent life of a patient.
Mayur Rahul, Narendra Kohli, and Rashi Agarwal
Inderscience Publishers
Anand Handa, Rashi Agarwal, and Narendra Kohli
Inderscience Publishers
Anand Handa, Rashi Agarwal, and Narendra Kohli
Springer Science and Business Media LLC
The spoken keyword recognition and its localization are one of the fundamental aspects of speech recognition and known as keyword spotting. In automatic keyword spotting systems, the Lip-reading (LR) methods have a broader role when audio data is not present or has corrupted information. The available works from the literature have focussed on recognizing a limited number of words or phrases and require the cropped region of face or lip. Whereas the proposed model does not require the cropping of the video frames and it is recognition free. The proposed model is utilizing Convolutional Neural Networks and Long Short Term Memory networks to improve the overall performance. The model creates a 128-dimensional subspace to represent the feature vectors for speech signals and corresponding lip movements (focused viseme sequences). Thus the proposed model can tackle lip reading as an unconstrained natural speech signal in the video sequences. In the experiments, different standard datasets as LRW (Oxford-BBC), MIRACL-VC1, OuluVS, GRID, and CUAVE are used for the evaluation of the proposed model. The experiments also have a comparative analysis of the proposed model with current state-of-the-art methods for Lip-Reading task and keyword spotting task. The proposed model obtain excellent results for all datasets under consideration.
Mayur Rahul, Narendra Kohli, and Rashi Agarwal
Inderscience Publishers
Anand Handa, Rashi Agarwal, and Narendra Kohli
Ubiquity Press, Ltd.
This paper presents a comprehensive, highly defined and fully labelled video dataset. This dataset consists of videos related to 67 different subjects. The videos contain similar text and the text contains digits from 1 to 20 recited by 67 different subjects using the same experimental setup. This dataset can be used as a unique resource for researchers and analysts for training deep neural networks to build highly efficient and accurate recognition models in various domains of computer vision such as face recognition model, expression recognition model, speech recognition model, text recognition, etc. In this paper, we also train models related to face recognition and speech recognition on our dataset and also compare the results with the publically available datasets to show the effectiveness of our dataset. The experimental results show that our comprehensive dataset is more accurate than other dataset on which the models are tested.
Mayur Rahul, Narendra Kohli, Rashi Agarwal, and Sanju Mishra
Inderscience Publishers
Mayur Rahul, Pushpa Mamoria, Narendra Kohli, and Rashi Agrawal
Springer Singapore
Partition-based feature extraction is widely used in the pattern recognition and computer vision. This method is robust to some changes like occlusion, background, etc. In this paper, partition-based technique is used for feature extraction and extension of HMM is used as a classifier. The new introduced multistage HMM consists of two layers. In which bottom layer represents the atomic expression made by eyes, nose, and lips. Further upper layer represents the combination of these atomic expressions such as smile, fear, etc. Six basic facial expressions are recognized, i.e., anger, disgust, fear, joy, sadness, and surprise. Experimental result shows that proposed system performs better than normal HMM and has the overall accuracy of 85% using JAFFE database.
Madhu Jadon, Rashi Agarwal, and Raghuraj Singh
IEEE
Manual leaf area estimation is a slow and tiresome process and it is also subjective and inaccurate. A new method is presented in this paper to make this process easier, user friendly and accurate. This paper proposes a new automatic and easy method for leaf area estimation based on digital images of leaves. Researchers can use this method very easily. A standard five rupee coin is used as a reference object to calculate the leaf area. Forty five leaves each of three pulses (i.e. Green Gram, Black gram and Pigeon Pea) have been taken for this study. The Results of our method are comparable to the results of grid paper method. Average accuracy of our method is approximately 97%. Results show that the proposed method is more accurate and user friendly.
Anand Handa, Rashi Agarwal, and Narendra Kohli
IEEE
Face Recognition is one of the most demanding problems in computer vision and image processing. Face recognition techniques can be divided into following categories: methods which are applicable on intensified images; images derived from video sequences; and images that require three-dimensional information and data. In this paper, we have compared the various Bi-Modal and Multi-Modal techniques. Speech recognition is an essential component for various applications such as building interfaces for natural human machines. Acoustic speech is used as the only input for various speech based automatic recognition system. Multimodal recognition is as an indigenous component of the next level speech-based systems. The main objective of this review paper is to compare the various components of bimodal recognition and it aims at some important ongoing research issues in the field of image and face recognition.