Convolutional Neural Network (CNN) and GMM Supervectors in Video Concept Detection.

Apply

Convolutional Neural Network (CNN) and GMM Supervectors in Video Concept Detection.

Journal Title: International Journal of Engineering and Science Invention - Year 2018, Vol 7, Issue 8

Abstract

Video concept detectionplays an important role in digital image processing. Video concept detectionis defined as detecting the concepts that are present in the videoshots. Concept canbeanything of users’ interest. Video concept detectionisassigning the labels to the input videoshots. For getting the accuratevideoshots, different types of techniques are used. The basic approach of concept detectionis to use classification algorithm. There are manydifferent CNN structures for video or image classification likeAlexNet, Overfeat, VGG, Network-in-network, GoogLeNet, ResNet. In thispaperwe propose to use CNN and SVM, to build Concept classifierswhichpredict the relevance between images or videoshots and a given concept. DeepConvolutional neural networks (CNN) are a special type of feed-forward networks. DCNN performverywell on visual recognition tasks. We are usingAlexnet architecture pre-trained on the ImageNETdataset on TRECVID dataset and using GMM supervectors corresponding to apply on different types of audio and visualfeatures.

Authors and Affiliations

Ritika D Sangale, Nita S Patil, Sudhir D Sawarkar

Keywords

Shot Boundary Detection Keyframe Extraction Feature Extraction Deep CNN GMM Supervectors SVM Score Fusion.

A Numerical Solutions Of Initial Value Problems (Ivp) For Ordinary Differential Equations (Ode) With Euler And Higher Order Of Runge Kutta Methods Using Matlab

This Paper Mainly Presents Euler Method And 4 thorder Runge Kutta Method (RK4) For Solving Initial Value Problems (IVP) For Ordinary Differential Equations (ODE). The Two Proposed Methods Are Quite Efficient And Practica...

EP ID EP397999
DOI -
Views 67
Downloads 0