Scene Description from Images to Sentences

1,Computer Engineering LDRP-ITR Gandhinagar, India 2Prof & HOD, Information Technology LDRP-ITR Gandhinagar, India ---------------------------------------------------------------------***--------------------------------------------------------------------Abstract— People exchange their views using language, whether spoken, written, or typed. A notable amount of this language describes the environment around us, especially the visual scenario in our surroundings or depicted in images or video. Scene description aims to generate the sentences from given set of input images. It links the visual perception with the language space. All present approaches are purely found in supervised machine learning setup. However, owing to the dearth of training data, this seldom achieves desired accuracy. We present a model that uses “Distributed Intelligence” as the prevalent theme in the artificial intelligence literature. Rather than only relying on the training dataset (PASCAL VOC 2012 containing 11530 images, FLICKR8K, FLICKR30K & MSCOCO), we harness the power of internet in order to generate more precise sentences related to the images.