Taresh Sarvesh Sharan

Bio: Taresh Sarvesh Sharan is an academic researcher from Indian Institute of Technology (BHU) Varanasi. The author has contributed to research in topics: Feature (computer vision) & Segmentation. The author has an hindex of 2, co-authored 6 publications receiving 11 citations. Previous affiliations of Taresh Sarvesh Sharan include Northeastern University & Indian Institutes of Technology.

Encoder Modified U-Net and Feature Pyramid Network for Multi-class Segmentation of Cardiac Magnetic Resonance Images

Abstract: Cardiovascular diseases are leading cause of death worldwide. Timely and accurate detection of disease is required to reduce load on healthcare system and number of deaths. For this, accurate and f...

An Augmented Deep Learning Network with Noise Suppression Feature for Efficient Segmentation of Magnetic Resonance Images

Abstract: The segmentation of cardiac MR images requires extensive attention as it needs a high level of care and analysis for the diagnosis of affected part. The advent of deep learning technology has paved...

Denoising and Spike Removal from Raman Spectra using Double Density Dual-Tree Complex Wavelet Transform

Abstract: We aim to show the effectiveness of the double density dual-tree complex wavelet transform to denoise the Raman signal. A comparative study is carried out of the double density dual-tree complex wavelet transform with the discrete wavelet transform, dual-tree complex wavelet transforms, and Savitzky–Golay smoothing method to show its capability and effectiveness. Results show that denoising based on the double density dual-tree complex wavelet transform can improve the quantitative and qualitative analysis of the Raman signal.

Evaluation of Deep Learning Methods (DnCNN and U-Net) for Denoising of Heart Auscultation Signals

Abstract: Every year, around 17.9 million people die due to Cardiovascular Diseases which is 31% of global deaths. These numbers indicate the need for a system that should be sensitive to detect Heart Disease at an early stage. Heart sound signals can give information about heart damage at a much earlier stage. For proper information extraction from heart auscultation about a heart condition, it is required that the signals are free from noise so that improper classification as the normal abnormal situation can be eliminated. A number of denoising methods have been proposed for denoising of heart auscultation sounds, both in the time domain and frequency domain. Most of them suffer from one or more problems to properly denoise the heart sound. DnCNN and U-Net have been used previously as a state of art method to denoise images. In this paper, we try to evaluate the ability of DnCNN and U-Net to denoise the signal.

Segmentation of Brain Tumour in MR Images Using Modified Deep Learning Network

Abstract: This paper presents a modified segmentation network for brain tumour segmentation in Magnetic Resonance Images. The early detection of brain tumour is quite mandatory for planning the treatment. This work proposes a computer-based automatic approach for the segmentation of brain tumour. The network proposed in this paper effectively delineated the boundaries of the brain tumour region. Exceedingly good results were obtained when the trained network was fed with other datasets. The network also showed a good improvement in the results when it was tested on real-time MRI datasets. An improvement of 7.6% and 7% was observed in the mIoU and BF score when the real time MR dataset of brain tumour was applied to the network. The network was incorporated using depthwise separable convolution.

Encoder Modified U-Net and Feature Pyramid Network for Multi-class Segmentation of Cardiac Magnetic Resonance Images

Abstract: Cardiovascular diseases are leading cause of death worldwide. Timely and accurate detection of disease is required to reduce load on healthcare system and number of deaths. For this, accurate and f...

Computer-aided automatic approach for denoising of magnetic resonance images

Abstract: This paper proposes a dual path deep convolution network based on discriminative learning for denoising MR images. The noise in MR images causes problems in identifying the regions of interest. The...

A Vehicle Detection Model Based on 5G-V2X for Smart City Security Perception

Abstract: Security perception systems based on 5G-V2X have become an indispensable part of smart city construction. However, the detection speed of traditional deep learning models is slow, and the low-latency characteristics of 5G networks cannot be fully utilized. In order to improve the safety perception ability based on 5G-V2X, increase the detection speed in vehicle perception. A vehicle perception model is proposed. First, an adaptive feature extraction method is adopted to enhance the expression of small-scale features and improve the feature extraction ability of small-scale targets. Then, by improving the feature fusion method, the shallow information is fused layer by layer to solve the problem of feature loss. Finally, the attention enhancement method is introduced to increase the center point prediction ability and solve the problem of target occlusion. The experimental results show that the UA-DETRAC data set has a good detection effect. Compared with the vehicle detection capability before the improvement, the detection accuracy and speed have been greatly improved, which effectively improves the security perception capability based on the 5G-V2X system, thereby promoting the construction of smart cities.