arXiv subjects

Shubham Jain

Publications and source records attributed to Shubham Jain.

57 records · Page 4Linked to original sources

arXiv2017

2D-3D Fully Convolutional Neural Networks for Cardiac MR Segmentation

In this paper, we develop a 2D and 3D segmentation pipelines for fully automated cardiac MR image segmentation using Deep Convolutional Neural Networks (CNN). Our models are trained end-to-end from scratch using the ACD Challenge 2017 dataset comprising of 100 studies, each containing Cardiac MR images in End Diastole and End Systole phase. We show that both our segmentation models achieve near state-of-the-art performance scores in terms of distance metrics and have convincing accuracy in terms of clinical parameters. A comparative analysis is provided by introducing a novel dice loss function and its combination with cross entropy loss. By exploring different network structures and comprehensive experiments, we discuss several key insights to obtain optimal model performance, which also is central to the theme of this challenge.

cs.CV↗

arXiv2017

SampleRNN: An Unconditional End-to-End Neural Audio Generation Model

In this paper we propose a novel model for unconditional audio generation based on generating one audio sample at a time. We show that our model, which profits from combining memory-less modules, namely autoregressive multilayer perceptrons, and stateful recurrent neural networks in a hierarchical structure is able to capture underlying sources of variations in the temporal sequences over very long time spans, on three datasets of different nature. Human evaluation on the generated samples indicate that our model is preferred over competing models. We also show how each component of the model contributes to the exhibited performance.

cs.SD↗

arXiv2013

A customized flocking algorithm for swarms of sensors tracking a swarm of targets

Wireless mobile sensor networks (WMSNs) are groups of mobile sensing agents with multi-modal sensing capabilities that communicate over wireless networks. WMSNs have more flexibility in terms of deployment and exploration abilities over static sensor networks. Sensor networks have a wide range of applications in security and surveillance systems, environmental monitoring, data gathering for network-centric healthcare systems, monitoring seismic activities and atmospheric events, tracking traffic congestion and air pollution levels, localization of autonomous vehicles in intelligent transportation systems, and detecting failures of sensing, storage, and switching components of smart grids. The above applications require target tracking for processes and events of interest occurring in an environment. Various methods and approaches have been proposed in order to track one or more targets in a pre-defined area. Usually, this turns out to be a complicated job involving higher order mathematics coupled with artificial intelligence due to the dynamic nature of the targets. To optimize the resources we need to have an approach that works in a more straightforward manner while resulting in fairly satisfactory data. In this paper we have discussed the various cases that might arise while flocking a group of sensors to track targets in a given environment. The approach has been developed from scratch although some basic assumptions have been made keeping in mind some previous theories. This paper outlines a customized approach for feasibly tracking swarms of targets in a specific area so as to minimize the resources and optimize tracking efficiency.

cs.OH↗