For Project Enquiry +91 9962 067 067

EXISTING SYSTEM

PROPOSED SYSTEM

EXISTING CONCEPT:

          The performance of tag matching-based retrieval approaches highly depends on the availability and the quality of manual tags. However, in practice, keywords/tags are expensive to obtain and often ambiguous.

         The semantic gap that occurred between the low-level features and the high-level semantic content and leads to Poor CBIR performance.

PROPOSED CONCEPT:

          Recent studies have shown that the content of the RS data is more relevant than manual tags. so we choose content of RS data instead of  manual tags.

          In order to confine the semantic gap, relevance feedback (RF) schemes have been designed to iteratively improve the performance of CBIR.

EXISTING  TECHNIQUE :

         SUPPORT VECTOR MACHINES (SVM)

PROPOSED ALGORITHM:

         TRIPLE CRITERIA ACTIVE LEARNING (TCAL)

TECHNIQUE DEFINITION:

          The Support Vector Machines (SVM) is an approximate implementation of the structural risk minimization (SRM) principle. The SVM can provide a good generalization performance on pattern classification problems without incorporating problem domain knowledge.

ALGORITHM DEFINITION:

         The proposed TCAL method is defined in the context of binary SVM classification and selects a batch S = {X1,X2, . . . ,Xh} of h images at each RF iteration that are as follows: 1) uncertain (i.e., ambiguous); 2) as more diverse as possible to each other; and 3) located in the highest density regions of the image feature space.

DRAWBACKS:

          The performance of tag matching-based retrieval approaches highly depends on the availability and the quality of manual tags. However, in practice, keywords/tags are expensive to obtain and often ambiguous.

          Images that fall into the high-density regions of the image feature (descriptor) space are crucial for CBIR problems particularly when a small number of initially annotated images are available.

ADVANTAGES:

          The annotation cost is reduced due to the avoidance of redundant images.

          Accurate retrieval accuracy can be obtained due to the improved class models estimated on a high-quality training set on the basis of the classification rule used from the considered classifier.

          The HI kernel has the advantage to be parameter-free.