CrowdSearch: Exploiting Crowds for Accurate Real-Time Image Search on Mobile Phones Original work by Yan, Kumar & Ganesan Presented by Tim Calloway.

Slides:



Advertisements
Similar presentations
Project Supervisor: Dr. Sanath Jayasena Project Coordinator: Mr. Shantha Fernando Athukorala A.U.B Dissanayake C.P. Kumara M.G.C.P. Priyadarshana G.V.J.
Advertisements

W3C Workshop on Web Services Mark Nottingham
Incentivize Crowd Labeling under Budget Constraint
CrowdER - Crowdsourcing Entity Resolution
Imbalanced data David Kauchak CS 451 – Fall 2013.
Yasuhiro Fujiwara (NTT Cyber Space Labs)
CrowdSearch: Exploiting Crowds for Accurate Real-time Image Search on Mobile Phones Original work by Yan, Kumar & Ganesan Presented by Shibo Li & Jian.
VTrack: Accurate, Energy-Aware Road Traffic Delay Estimation Using Mobile Phones Arvind Thiagarajan, Lenin Ravindranath, Katrina LaCurts, Sivan Toledo,
Contextual Advertising by Combining Relevance with Click Feedback D. Chakrabarti D. Agarwal V. Josifovski.
Carnegie Mellon 1 Maximum Likelihood Estimation for Information Thresholding Yi Zhang & Jamie Callan Carnegie Mellon University
What is adaptive web technology?  There is an increasingly large demand for software systems which are able to operate effectively in dynamic environments.
University College London Department of Computer Science CS M038/GZ06: Mobile and Cloud Computing Paper presentation  Students: Shaig Mursalzade & Vasos.
Niranjan Balasubramanian Aruna Balasubramanian Arun Venkataramani University of Massachusetts Amherst Energy Consumption in Mobile Phones: A Measurement.
ALBERT PARK EEL 6788: ADVANCED TOPICS IN COMPUTER NETWORKS Energy-Accuracy Trade-off for Continuous Mobile Device Location, In Proc. of the 8th International.
H-1 Network Management Network management is the process of controlling a complex data network to maximize its efficiency and productivity The overall.
Designing Ranking Systems for Hotels on Travel Search Engines by Mining User-Generated and Crowd sourced Content Author - Anindya Ghose, Panagiotis G.
What is Software Architecture?
Presented by: Michal Nir, Saar Gross Supervisors: Nadav Golbandi, Oren Somekh Computer Science Department Industrial Project (234313) Tuesday, January.
Wang, Z., et al. Presented by: Kayla Henneman October 27, 2014 WHO IS HERE: LOCATION AWARE FACE RECOGNITION.
Chapter 10: Authentication Guide to Computer Network Security.
A User Experience-based Cloud Service Redeployment Mechanism KANG Yu.
A Wanderer’s Guide to the Data Network Navigating Our Connected World Garrett Shields, GISP, CFM.
Micro-Blog : Sharing and Querying Content Through Mobile Phones and Social Participation Presented by: Muhammad S. Karim By S. Gaonkar, J. Li, R. Choudhury,
AS Level ICT Mrs. Ghazaal. In the past, when a customer wanted to talk to someone in a company they would usually be able to telephone and be put through.
Introduction to Computers. Objectives Overview Describe the five components of a computer Discuss the advantages and disadvantages that users experience.
Active Learning for Class Imbalance Problem
UNIVERSITY of NOTRE DAME COLLEGE of ENGINEERING Preserving Location Privacy on the Release of Large-scale Mobility Data Xueheng Hu, Aaron D. Striegel Department.
WSN Done By: 3bdulRa7man Al7arthi Mo7mad AlHudaib Moh7amad Ba7emed Wireless Sensors Network.
 An important problem in sponsored search advertising is keyword generation, which bridges the gap between the keywords bidded by advertisers and queried.
Lecture 9: Chapter 9 Architectural Design
How to Integrate a WiFi for Mobile Networks Karen Cameron, EdTech 541 Image from wifinotes.com.
UOS 1 Ontology Based Personalized Search Zhang Tao The University of Seoul.
CIKM’09 Date:2010/8/24 Advisor: Dr. Koh, Jia-Ling Speaker: Lin, Yi-Jhen 1.
Chapter 13 Architectural Design
1 Distributed Energy-Efficient Scheduling for Data-Intensive Applications with Deadline Constraints on Data Grids Cong Liu and Xiao Qin Auburn University.
Content Sharing over Smartphone-Based Delay- Tolerant Networks.
Jan. 29, 2002Grand Challenges in Simulation Issues in Enhancing Model Reuse C. Michael Overstreet Richard E. Nance Osman Balci.
CSM06 Information Retrieval Lecture 6: Visualising the Results Set Dr Andrew Salway
1 A Systems Architecture for Ubiquitous Video Neil J. McCurdy and William G. Griswold Mobisys, 2005 Presented by Sangjae Lee.
Mobile Agent Migration Problem Yingyue Xu. Energy efficiency requirement of sensor networks Mobile agent computing paradigm Data fusion, distributed processing.
CrowdSearch: Exploiting Crowds for Accurate Real-Time Image Search on Mobile Phones Original work by Tingxin Yan, Vikas Kumar, Deepak Ganesan Presented.
A Study of Smartphone User Privacy from the Advertiser's Perspective Yan Wang 1, Yingying Chen 1, Fan Ye 2, Jie Yang 3, Hongbo Liu 4 1 Department of Electrical.
Cmpe 589 Spring 2006 Lecture 2. Software Engineering Definition –A strategy for producing high quality software.
Human Tracking System Using DFP in Wireless Environment 3 rd - Review Batch-09 Project Guide Project Members Mrs.G.Sharmila V.Karunya ( ) AP/CSE.
1.Research Motivation 2.Existing Techniques 3.Proposed Technique 4.Limitations 5.Conclusion.
TEMPLATE DESIGN © E-Eye : A Multi Media Based Unauthorized Object Identification and Tracking System Tolgahan Cakaloglu.
A BRIEF INTRODUCTION TO CACHE LOCALITY YIN WEI DONG 14 SS.
Hand Gesture Recognition Using Haar-Like Features and a Stochastic Context-Free Grammar IEEE 高裕凱 陳思安.
 An Information System (IS) is a collection of interrelated components that collect, process, store, and provide as output the information needed to.
SAPIR Search in Audio-Visual Content using P2P Information Retrival For more information visit: Support.
Privacy Issues in Smart Grid R. Newman. Topics Defining anonymity Need for anonymity Defining privacy Threats to anonymity and privacy Mechanisms to provide.
Web Technologies Lecture 13 Introduction to cloud computing.
REU 2007 Computer Science and Engineering Department The University of Texas at Arlington Research Experiences for Undergraduates in Information Processing.
Identifying “Best Bet” Web Search Results by Mining Past User Behavior Author: Eugene Agichtein, Zijian Zheng (Microsoft Research) Source: KDD2006 Reporter:
Power Guru: Implementing Smart Power Management on the Android Platform Written by Raef Mchaymech.
Multimedia Retrieval Architecture Electrical Communication Engineering, Indian Institute of Science, Bangalore – , India Multimedia Retrieval Architecture.
1 Traffic Engineering By Kavitha Ganapa. 2 Introduction Traffic engineering is concerned with the issue of performance evaluation and optimization of.
Device and Application Engineering Design 6. Overview EZ-Park Device Displays validity of parking spot Facilitates payment for parking Main Components.
Resource Optimization for Publisher/Subscriber-based Avionics Systems Institute for Software Integrated Systems Vanderbilt University Nashville, Tennessee.
LOGO Supervisor: Mr.Huỳnh Anh Dũng Students: Nguyễn Công Tuyến Nguyễn Cảnh Phương Phạm Thị Hằng Bùi Thị Huệ Trần Đức Bình Nguyễn.
SOURCE:2014 IEEE 17TH INTERNATIONAL CONFERENCE ON COMPUTATIONAL SCIENCE AND ENGINEERING AUTHER: MINGLIU LIU, DESHI LI, HAILI MAO SPEAKER: JIAN-MING HONG.
Huber Flores Social-aware Hybrid Mobile Offloading A contribution for edge and fog computing? Huber Flores
Objectives Overview Explain why computer literacy is vital to success in today’s world Define the term, computer, and describe the relationship between.
Fundamental of Information Communication Technology (ICT)
Improving searches through community clustering of information
Erasmus University Rotterdam
Cloud Computing By P.Mahesh
Visible Light based Activity Sensing using Ceiling Photosensors
th IEEE International Conference on Sensing, Communication and Networking Online Incentive Mechanism for Mobile Crowdsourcing based on Two-tiered.
Akshay Tomar Prateek Singh Lohchubh
Presentation transcript:

CrowdSearch: Exploiting Crowds for Accurate Real-Time Image Search on Mobile Phones Original work by Yan, Kumar & Ganesan Presented by Tim Calloway

Roadmap  Problem Description  What is “crowdsourcing”?  System Architecture  The CrowdSearch Algorithms  Delay Prediction  Validation Prediction  Experimental Evaluation  Discussion/Criticism  Questions

The Perceived Problem  Text-based search is easy…

The Perceived Problem  Mobile-based search will become more important in the future.  More than 70% of smart phone users perform searches.  Expected to be more mobile searches than non-mobile searches soon  Text-based mobile searches are easy as well…

The Perceived Problem  But how does a mobile phone user search for this?  No visible words/letters; too far away to know the address.

The Perceived Problem  Ways to find out what that building is:  Ask random people on the street  Travel to the building to see the address/sign  Take a picture of the building with your mobile device and send to a search engine…  How easy is image searching on a mobile phone though?

The Perceived Problem  Image search is a non-trivial problem – have to deal with variations in lighting, texture, image quality, etc.  Even when results are returned, scrolling through multiple pages on a mobile device is cumbersome.  Search should be precise and return very few erroneous results.  Multimedia searches require significant  Memory  Storage  Computing resources

The Proposed Solution  CrowdSearch – Attempts to provide an accurate, image search system for mobile devices by combining…  Automated image search and  Real-time human validation of search results  Leverage crowdsourcing through Amazon Mechanical Turk (AMT)

The Proposed Solution  Humans are good at comparing images  Could an automated search determine these two images are of the same building?  Crowdsourcing increases search result accuracy.

Balancing Tradeoffs  Result delay  Should minimize delay or at least keep it within a user-provided bound  Result accuracy  Strive for high (i.e., ≥ 95%) accuracy  Monetary cost  Low cost is better than high cost  Energy  Should consume minimal battery power

System Architecture

 Three main components:  Mobile Device  Initiates queries  Displays responses  Performs local image processing (maybe)  Remote Server  Performs automated image search  Triggers image validation tasks  Crowdsourcing System (AMT)  Validates image search results

System Operation Overview

 How do we minimize delay and cost while maximizing accuracy?

Accuracy Considerations  How many validations are required for 95% accuracy?  Requiring at least three validations out of five achieves ≥ 95% accuracy.

Optimizing Delay  Utilize parallel posting  Post all candidate images to the crowdsourcing system at the same time.  But this approach increases cost! 5 cents = 20 cents 5 cents

Optimizing Cost  Utilize serial posting  Post top-ranked candidate first, wait for responses, then post next candidate if necessary.  This approach increases delay!

CrowdSearch Delay/Cost Optimization  Combine elements of parallel and serial posting  Prediction requires delay and validation models  Goal: want at least one verified result by the deadline.

CrowdSearch Delay/Cost Optimization

Delay Prediction Model  The delay of a single response is the combination of acceptance delay and submission delay.  Both of these follow an exponential distribution with an offset.  Thus, overall delay is the convolution of these delays.

Delay Prediction Model Performance

Validation Model  Given a response set S, want to compute probability of positive validation result.  Use training data to set these probabilities  If the probability of a positive result is less than some threshold, send the next candidate to validation.  In this example, if the threshold were set to < 76%, the server would post the next candidate image to AMT.

Power Considerations  Should some image processing occur on the local device or should it be outsourced to the server?  It depends!  Use remote processing when WiFi is available.  Use local processing when only 3G is available

Experimental Results  Any of the crowdsourcing schemes lead to better results!  Some types of images are easier for automated searches to handle than others

Experimental Results  CrowdSearch leads to (given a long enough deadline)…  Behavior close to parallel posting for recall  Behavior close to serial posting for search cost

Thoughts/Criticism  The limited nature of the solution  Limitation to the four categories  Buildings  Books  Flowers  Faces  Only 1000 images in the backend database.  Would increasing the number of automated search images increase total task time in a significant way?

Thoughts/Criticism  How useful is this anyway?  Are people willing to go through the trouble to set up a payment account and pay 5-20 cents for a search?  How much effort would it usually take for someone to find out what the object is through traditional means?  Especially for books!  Privacy concerns  People utilizing CrowdSearch must accept the fact that random strangers know what they are looking at and searching for.  Additionally, their GPS information might be provided to the CrowdSearch servers.  What about the privacy of the object of the search?  Undercover police officers

Questions?