Nov 2021 – Apr 2022 · 5 mos · Bengaluru, Karnataka, India · Remote

• Proposed Model :
• The proposed algorithm includes Context based Image captioning with Sentiment analysis . The landscape images with their conceptual captions are loaded in the proposed algorithm . The algorithm will initially extract all colour features of the model and understand the colour psychology based moods that can be predicted using a custom feature extractor script developed using Freud's analysis . Where each colour is mapped with an emotion with proper weighted values . Hence colours are tagged accordingly . Then the proposed will be detecting the major object out of the Image and find out the context vectors of the background excluding the vectors of the image . Based on the above feature extracted by VGG 16 , the connected LSTM node will generate contextual and conceptual image caption of the landscape imagery . Next the Captions are passed into the sentiment analysis model CU DNN LSTM with 5 fold attention , where the text are classified into seven emotion classes and one class is marked as output . The model is judging the inter relationship between word emotion vectors to judge the expected mood .