Content Based Image Search - Computer Vision | EECS 841, Study notes of Electrical and Electronics Engineering

Download Content Based Image Search - Computer Vision | EECS 841 and more Study notes Electrical and Electronics Engineering in PDF only on Docsity!

Content Based Image Search

By

Aravindkumar Ilangovan

Content

 Introduction

 Problem

 Solution

 Implementation

 Results

Problem

 web based image search is based on

tag information associated with

images.

Solution

 Content based image search.

Implementation

Object

Feature

Extraction

Object

Feature

Learning

Object Detection

Object training

Database

Search

Space

Result

Threshold = 10 Threshold = 15 Threshold = 20 Threshold = 25 Threshold = 30 P N P N P N P N P N Training Sample Car 100 0 98 2 77 23 55 45 28 72 NonCar 33 67 26 74 16 84 7 93 0 100 True and False positive rates 0.75 0 0.79 0.02 0.82 0.21 0.88 0.32 1 0. Accuracy 0.835 0.86 0.805 0.74 0. Threshold = 10 Threshold = 15 Threshold = 20 Threshold = 25 Threshold = 30 P N P N P N P N P N NonTraining Sample car 100 0 100 0 87 13 38 62 3 97 NonCar 96 4 81 19 50 50 20 80 0 100 True and False positive rates 0.51 0 0.55 0 0.63 0.20 0.65 0.43 1 0. Accuracy 0.52 0.595 0.685 0.59 0.

Future Work and Conclusion

 More images for learning.

 More Objects.

 As a filter for the current image

search.

Question?