Upload
edana
View
37
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Statistical techniques for video analysis and searching chapter 9. 2010-11806 Anton Korotygin. Contents. Introduction Model Vectors Video Search Fusion Experiments Conclusion. Before Starting. - PowerPoint PPT Presentation
Citation preview
Statistical techniques for video analysis and searching
chapter 9
2010-11806 Anton Korotygin
Contents
• Introduction • Model Vectors • Video Search Fusion • Experiments • Conclusion
Before Starting
• What is the main approaches for video analysis based on vector model indexing and interactive search fu-sion ?
• Which technique we apply in this ap-proaches ?
• What detectors we can use for that technique ?
Basic problems
Solution
• CBR – Content – based retrieval –searching and matching through the video based on similarity of its content
• MBR – Model – based retrieval – searching based on automatically extracted labels and detection results
• TBR – Text-based retrieval – applies to tex-tual forms of information related to the video which includes transcripts, embed-ded, text, speech, metadata, etc…
How it works ?
Approaches Techniques
• Model Vector
Model Vectors
• Priori learning of detectors • Concept detection and score map-
ping to produce model vectors
Priori learning of detectors
Concept learning
• Detection techniques – Support Vector Machines (SVM)– Gaussian Mixture Models (GMM)– Hidden Markov Models (HMM)
Concept detection
Support Vector Machines
Gaussian Mixture Models
Model Vector Construction
Model Vector Retrieval
Q&A
Thank you!!!