Youtor
Bag Of WordsBag of Visual Words based Image Recognition
>> YOUR LINK HERE: ___ http://youtube.com/watch?v=uIWRHlOgzBU
Welcome to all • This video demonstrates MATLAB implementation of Image Classification using Bag of Visual Words (BoVW). • Bag of Visual Words: In Computer Vision, object detection is being always main task/problem. In this project, image is classified using Bag of Visual Words (BoVW) which is also known as Bag of Features. • BoVW works on the principle of Bag of Words where frequency/occurrence of any word in a document is calculated. Likewise, in this technique, image is considered as sentence/group of words test image (which is being classified) is considered as document. Later according to frequency/occurrence of the visual word, that image is classified. • First, a classifier is designed with any public dataset (Caltech 101 in our case which is discussed below). For evaluating the classification, dataset is partitioned into Training Images and Valid Images (70% for Training purpose and 30% for Evaluation). Bag of Features of each training image are extracted as the procedure is shown in above pictures and classifier using SVM is designed. Accuracy is calculated using Valid Images. • Test images are downloaded from Google. • About the datasets: Caltech 101 dataset is used for training purpose. • In this dataset, pictures of objects belonging to 101 categories are stored. About 40 to 800 images per category. Most categories have about 50 images. Collected in September 2003 by Fei-Fei Li, Marco Andreetto, and Marc 'Aurelio Ranzato. The size of each image is roughly 300 x 200 pixels. • This dataset may be downloaded from http://www.vision.caltech.edu/Image_D... • Reference papers: • Many papers on this topic is available on internet. • First, train the classifier, then download the test images from Google which would not be part of Training Images or Valid Images. • Thank You for watching my video. • Feel Free to Contact at • [email protected]
#############################
