BHUPENDRA D FATANIYA
@nirmauni.ac.in
Assistant Professor
Nirma University
RESEARCH INTERESTS
Computer Vision, Image Processing, Machine Learning, Artificial Intelligence, Embedded Systems
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Rutul Patel, Neel Patel, Bhupendra Fataniya, and Dhaval Shah
Springer Science and Business Media LLC
Yash Jha, Harsh Prajapati, and Bhupendra Fataniya
IEEE
Object detection has been evolving greatly in recent years and the advancements in hardware and software technologies have made it possible to perform object detection with ease. Due to the enhanced capabilities of the modern processors and Graphics Processing Unit (GPU) of doing an exponentially complex and extensive number of iterations in very less time. Real-time object detection has become highly popular and the center of attention in recent years because most of the hardware owned by common users is powerful enough to compute that which unlocks whole new possibilities for implementing real-time object detection in numerous applications in various domains. Real-time herbal plant detection is one such topic that has many applications in the field of ayurvedic medicines and many other pharmaceutical applications that can be used to spike the efficiency in identifying these herbal plants that can be used as a precaution and even as a cure for numerous health problems. There are many existing algorithms for real-time detection, but the evolution of new Artificial Neural Network (ANN) and Machine Learning (ML) techniques unlocks new ways to implement recent and advanced algorithms to apply for real-time detection of such powdered microscopic images to achieve better performance in various aspects compared to already existing methods. Our model is trained for detecting three types of microscopic herbal plants.
Rohan Malhotra, Hemang Patel, and Bhupendra D. Fataniya
IEEE
With the rising of the new pandemic, problems to detect the presence of Covid-19 also emerged. To track the infections, RT-PCR and rapid testing are followed in the current situation which is time-consuming and could be an important time for severe patients. To decrease the amount of time for COVID-19 prediction, Chest X-rays could play an important role in determining the result. So by using Chest X-rays with Artificial Intelligence, the COVID-19 disease can be detected in a lesser amount of time under the guidance of the specialist. For this, the Deep Learning techniques like Convolutional Neural Networks (CNN) have been proved quite successful for image recognition and classification. In this experiment, Covid-19 was detected with the help of ResNet architecture whose accuracy increases while going into deeper layers by using skip connections. ResNet is a pre-trained model on the ImageNet database. During the experiment, ResNet18 architecture was used because it has the least number of layers as compared to other CNN architectures and so, for determining the best accuracy obtained with lesser computations. Methods like k-fold cross-validation, confusion matrices, etc were used in obtaining the accuracy of around 89% for COVID-19 prediction. Hence, CNN could be a useful tool for the prediction of COVID-19 and saving time for both patients and doctors for further treatment.
Riddhi Soni, Sachin Gajjar, Manisha Upadhyay, and Bhupendra Fataniya
Springer Singapore
Rohan Marwaha and Bhupendra Fataniya
IEEE
The objective of this paper is to proficiently classify the microscopic images of powder of Indian Herbal plants. Since they hold great importance in medicine industry and their identification is only done by experts for the powdered form, we have eluded the need for an expert by automating the process, yielding decent results. Although, attempts have been made to perform this task but the methodologies used do not provide the results with high accuracy. Inspired from the state-of-the-art deep learning techniques we have performed the classification by fine-tuning 4 pre-trained models provided by the Keras library which have provided with great results on ImageNet dataset. Out of the 4 models used, VGG16 provides the highest accuracy, Precision, Recall and F1 score but is the slowest to train. MobileNet is fastest but is mediocre in other parameters while Xception is 2nd fastest but with lowest accuracy and InceptionV3 with mediocre results.
Bhupendra Fataniya, Tanish Zaveri, and Sanjeev Acharya
American Scientific Publishers
Shraddha Vyas, Bhupendra Fataniya, Tanish Zaveri, and Sanjeev Acharya
ACM
This paper proposes an automated algorithm for plant identification using microscopic images of powder of herbal plants. In current scenario, the task of identifying plant from its powder form is done by pharmaceutical companies, which perform this task manually. This process takes lots of effort and time. Microscopic image of powder contains varieties of information, which are important evidence for identification of the plant. With every image, different type of noise are present, which makes the segmentation as a critical job. In this paper, we are proposing an algorithm which performs this task automatically by a computer. Our method consists two steps: "Pre-Processing" and "Image Segmentation". Firstly, microscopic images of "Liquorice" and "Rhubarb" plants were taken. On those images Top-hat and Bot-hat transformation are performed. Wiener Filter is used for image smoothing. An image segmentation is performed using Otsu's thresholding algorithm and find region of interest. The extra blobs were removed using morphological operations. Our proposed algorithm shows the efficiency for successfully detection of Liquorice and Rhubarb plants are 91.37% and 92.94% respectively.
A. I. Mecwan, D. G. Shah, and B. D. Fataniya
IEEE
Education in India in 21st century demands innovations in the evaluation methods with the emergence of Outcome Based Education (OBE). Lots of Research is already done in innovations in teaching and learning methods, but very few literatures on the innovations in evaluations are available. The paper discusses various innovative methods for the evaluation of students. The pros and cones of traditional exam and the new methods are also discussed. A survey on the various method is carried out and the results of the same are also presented.
Bhupendra Fataniya, Meet Joshi, Urmil Modi, and Tanish Zaveri
Elsevier BV
Bhupendra Fataniya, Prachi Manishkumar Patel, Tanish Zaveri, and Sanjeev Acharya
IEEE
Identification of medicinal plant from its powder is a common task performed in pharmaceutical labs and industries. In this paper, we are proposed a identification using machine intelligence. The algorithm is designed for the microscopic images of plant powder and segment the specific object from these images. From the extracted image unique features are identified and using this we train our algorithm and classify the test images. We have prepared the image database at our pharmacy lab for two plants namely Licorice and Kurchi plant. The proposed algorithm is applied on the variety of fifty images and results are described using confusion matrix and ROC curves.
Priyank Chaudhary and Bhupendra Fataniya
IEEE
Super-Resolution is the process of constructing a high resolution image when a set of one or more low resolution input images is given. Traditionally, there are two methods exploited widely for enhancing the image via Super-Resolution viz. Single-Frame or Single-Based approach and Multi-Frame or Sequence-Based approach. Because the low resolution images have less information because of lower pixel density than their high resolution counterparts, the enhancement process requires missing image data to be calculated. In this paper, we have proposed a novel method that exploits the advantages of both these traditional methods. In the first phase, we improve a set of low resolution images via learning dictionary single frame method and in second phase we combine these by projecting these images onto convex sets thereby enhancing the image by information procured from multiple images. Experimental results show that our method works considerably better than state-of-the art Super Resolution enhancement methods.