Bag of Words (BoW)

Based on:
[**Object Retrieval with Large Vocabularies **]

Requirements

• MATLAB
• VLFeat (MATLAB API)

Usage Instructions

$ cd BoWImageMatching $ matlab

In MATLAB:
```matlab
>> cd src
>> edit bow_getDefaultParams.m % change the path to vlfeat's vl_setup script
>> bow_getDefaultParams; % get the params variable with default settings

Learn Vocabulary

>> model = bow_computeVocab('~/imagesDir', params);

Alternate option: Train only on subset of images from imagesDir/, given from a list

>> model = bow_computeVocab('~/imagesDir', params, 'imgsListFpath', '/path/to/TrainSet.txt');

Learn Inverted Index over corpus

>> iindex = bow_buildInvIndex('~/imagesDir', model);

Alternate option: Train only on subset of images from imagesDir/, given from a list

Test time: search for a query image

>> 
Note: Geometric reranking done by fitting a fundamental matrix using RANSAC, and counting the number of inliers.