An efficient framework for classifying the Clothing items based on fashion and fabric of the images

In recent days classification of images over fashion domain has become fundamental research problem with lot of computer vision based applications. In most of the existing research image classification is done based on the labels on the objects, however in the real world scenarios the images needs to classified into the labels based on the domain specific with proper guidelines, therefore the existing research works fails in achieving the evaluation measures like classification accuracy, precision, recall and time take taken for classification, Hence there is need of an efficient frameworks that classifies the images based on fashion and fabric by addressing existing research problems. This paper presents an efficient framework for segmenting using grab cut techniques with the integrating the histograms of oriented gradients (HOG) and the speeded-up robust features (SURF) techniques. Whereas HOG technique is used to retrieve the global features and SURF technique is used for the local features and then segmentation is done based on Scale-Invariant Feature Transform(SIFT) method that segments the line, solve the incident scaling, changes in lighting condition and rotation between the two images, Later classification of the clothing images is done by using feature matching classification techniques, with proper text analysis the classification of images is done based on color and fabric of the image. The proposed technique is compared with existing multi-level classification technique to prove the proposed framework is more efficient than existing works. With the proposed technique, we could able to achieve the accuracy of 93% by varying the dataset of the images which is 10–15% more accurate than existing multi-level classification.