Download - Inside proposal 16 113 - version 01
VIDEO SEARCHING BY AUTOMATIC ANNOTATION
GROUP NUMBER : 16 - 113
INSIDEPROJECT PROPOSAL
Group Members ,
IT 13122942 Wickramasinghe K.U
IT 11150558 Ashangani S.K
IT 13115494 De Silva D.W.N
IT 13112424 Gamwara V.M
INTRODUCTION
INSIDE
What is video
Sequence of images to form a moving picture
Our Mission
Friendly, simple video searching using automatic annotation to provide an accurate result
RESEARCH PROBLEM
Available search engines only provide Keyword searching, Audio searching, Image searching.
Most videos are weakly labeled or have misleading names.
Less applications with automatic video annotation.
SOLUTION
“INSIDE” a smart semantic video searching application with automatic annotation which support easy, quick user friendly querying and return an accurate list of videos for the requesting query.
LITERATURE REVIEW
Image subtraction and histogram comparison, traditional video shot boundary detection techniques – video slicing
( reference : Video summarization by video structure analysis and graph optimization Shi Lu,Department of Computer Science and Engineering)
Object detection techniques: Appearance Based Methods, Geometry-Based Methods (reference: OBJECT RECOGNITION METHODS BASED ON TRANSFORMATION COVARIANT FEATURES Jiri Matasand Stepan Obdrzalek)
Key word searching vs Semantic searching (reference : https://www.searchenginejournal.com/seo-101-semantic-search-care/119760/)
OBJECTIVES
Allow semantic video searching by analyzing the structure and detecting content objects.
Classify videos automatically without user interaction.
User friendly
Accurate
METHODOLOGY WITH TOOLS AND
TECHNIQUES
Video structure analysis – Shot boundary detection and video slicing
Deep Learning - Tensorflow
Data set preparation
Textual searching - Semantic searching
SYSTEM OVERVIEW
FUNCTIONS OF MEMBERS
Member Components Task
De Silva D.W.N Video structure analysis Identify shot boundaries. Fragment video into shots.Fragment shot into frames.
Ashangani S.K Deep Learning Neural Network configurations
Gamwara V.M Data set preparation Creation manipulation of large data set to train the Neural Network
Wickramasinghe K.U Textual searching Take search query as an input and identify a relationship among the words, return the output
BENEFITS FOR USER
Accurate, Efficient search results.
Can be used at any time any where.
Can annotate video automatically.
COMMERCIAL VALUE Can be used as a tool and implement in search engines
Can be used to categorize videos without user interaction
Ability to rate videos according to their category (harmful)
Thank You