For Project Enquiry +91 9962 067 067

AUGMENTING IMAGE DESCRIPTIONS USING STRUCTURED PREDICTION OUTPUT

ABSTRACT:-

Automatic Image Annotation (AIA) has been extensively studied, image descriptions with the output labels lack sufficient information. The need for richer descriptions of images arises in a wide spectrum of applications ranging from image understanding to image retrieval. This paper proposes to augment image descriptions using structured prediction output. We define a hierarchical tree-structured semantic unit to describe images, from which we can obtain not only the class and subclass one image belongs to, but also the attributes one image has. After defining a new feature map function of structured SVM, we decompose the loss function into every node of the hierarchical tree-structured semantic unit and then predict the tree-structured semantic unit for testing images. In the experiments, we evaluate the performance of the proposed method on two open benchmark datasets and compare with the state-of-the-art methods.

EXISTING SYSTEM	PROPOSED SYSTEM
EXISTING SYSTEM: Methods of generating images’ descriptions have been moved from the image classification methods (i.e., the single label methods) to the methods of multi-label image annotation and region tagging with the visual saliency estimation techniques.	PROPOSED SYSTEM: We have presented a hierarchical tree structured semantic unit to describe images. The proposed structured image description method could jointly output the class, subclass and attributes for images, resulting in more informative structured image descriptions. Our method augmented structured SVM to solve the problem of image descriptions prediction. .
EXISTING TECHNIQUE Visual saliency estimation techniques.	PROPOSED TECHNIQUE Hierarchical tree structured semantic
TECHNIQUE DEFINITION Generating images’ descriptions have been moved from the image classification methods (i.e., the single label methods) to the methods of multi-label image annotation and region tagging.	TECHNIQUE DEFINITION According to the semantic hierarchy in Image Net, we choose the root category and a low-level subcategory to describe the sample image in different semantic levels
DRAWBACKS It describes output from the multi-label image annotation and region tagging more informative than that of the image classification. Semantic labels are associated with one image or each object in one image..	ADVANTAGES Compared with alternative structured prediction method CRF and other multi-label algorithms, our method has shown superior performance for image descriptions. Informative descriptions of images are important for either image understanding or image retrieval..