Query Modeling for Click Data Based Image Recognition Using Graph Based Propagation and Sparse Coding

Abstract

We address the fine-grained image recognition problem using user click data, wherein each image is represented as a semantical query-click feature vector. Usually, the query set obtained from search engines is large-scale and redundant. We propose a novel query modeling approach to merge semantically similar queries and construct a compact click feature. We represent each query as a click feature, and design a graph based propagation approach to predict the zero-clicks, ensuring similar images have similar clicks. Afterwards, using this feature, we formulate the problem as a sparse coding based recognition task, wherein the dictionary is discriminatively trained. We evaluate our method for fine-grained image recognition on the public Clickture-Dog dataset. It is shown that, the propagated click feature performs much better than the original one. Also, sparse coding performs better than K-means in query merging.