For Project Enquiry +91 9962 067 067

WEAKLY SUPERVISED MULTI-GRAPH LEARNING FOR ROBUST IMAGE RERANKING

ABSTRACT

Visual re-ranking has been widely deployed to refine the traditional text-based image retrieval. Its current trend is to combine the retrieval results from various visual features to boost reranking precision and scalability. And its prominent challenge is how to effectively exploit the complementary property of different features. Another significant issue raises from the noisy instances, from manual or automatic labels, which makes the exploration of such complementary property difficult. This paper proposes a novel image reranking by introducing a new Co-Regularized Multi- Graph Learning (Co-RMGL) framework, in which intra-graph and Inter-graph constraints are integrated to simultaneously encode the similarity in a single graph and the consistency across multiple graphs. To deal with the noisy instances, weakly supervised learning via co-occurred visual attribute is utilized to select a set of graph anchors to guide multiple graphs alignment and fusion, and to filter out those pseudo labeling instances to highlight the strength of individual features. After that, a learned edge weighting matrix from a fused graph is used to reorder the retrieval results. We evaluate our approach on four popular image retrieval datasets and demonstrate a significant improvement over state-of-the-art methods.

EXISTING SYSTEM	PROPOSED SYSTEM
EXISTING CONCEPT:- The existing text-based image search engines, visual reranking has been received increasing attention in recent years. Such instances are obtained in either unsupervised or supervised manner, referred to pseudo relevance feedback, or user specification, respectively, both of which in some cases are denoted as “query image”. Unsupervised reranking method directly learns a ranking model from automatically acquired training data.	PROPOSED CONCEPT:- Proposed a graph-based learning approach that adaptively integrates multiple types of features into the graph affinity matrix towards a flexible reranking. The selection of labeling instances based on pseudo relevance feedback is not always correct. On the other hand, exploiting every visual elements in every individual labeling instance is still away from capturing the essence of user intension in query.
EXISTING TECHNIQUE:- Text based image search technique.	PROPOSED TECHNIQUE:- Co-Regularized Multi-Graph Learning.
TECHNIQUE DEFINITION:- Google Image Search and Microsoft Bing Image Search, are built based upon text search techniques which rank images by the textual similarity between the query keywords and the image tags, such as, title, description, surrounding Text. However, text-based search alone is not enough, due to the well-known semantic gaps between textual description and image content.	TECHNIQUE DEFINITION:- Co-Regularization Multi-Graph Learning for visual reranking, in which multiple retrieval results using different visual features are fused based on a graph learning formulation. CR-MGL provides a novel perspective in graph based reranking by considering both Intergraph agreement and inter-graph consistency under the anchor-based, attribute-induced supervision.
DRAWBACKS:- Text-based search alone is not enough, due to the well-known semantic gaps between textual description and image content.	ADVANTAGES:- Graph-based approaches to capture the intrinsic manifold structure hidden in labeling instances and to discover the underlying semantic relationship.