Clothing Retrieval Based on Inverted Index and Classification

We propose a new approach in clothing retrieval which aims at retrieval similar clothing from a lot of images with different clothing types. In our method, we first classify the images into different categories according to texture features, and then label the same category images with the category information. Finally, we select the images with the same categories as candidates, and obtain the final re-ranked result based on the inverted index technology. Experiment results demonstrate that our method can not only improve retrieval accuracy, but also reduce the time computing of the full retrieval.