Paper Title
Image Caption Generation Using CNN-LSTM Encoder-Decoder Model

Abstract
The increased amount of uploaded digital images, which are uploaded to the Internet, in particular, to social media and multimedia. Automated systems have been demanded because of the availability of platforms capable of reading images and giving coherent results description. Image tagging is a manual job that is tedious when done manually. Transformations that exist between annotators and is not very scalable datasets. The systems that rely on rules which are traditional are ineffective. Find how to make description of pictures that is true, substantial and grammatically correct. The aim of the research is to design a system and develop an image captioning auto system by the visual processing and text generation procedure. The system is based on the principle of an encoder-decoder in accordance with which a Convolutional Neural Network (CNN) functions as a construction, recognizes important visual patterns of images, and an LSTM network creates extracts of these patterns. The model was trained by the Flickr8k dataset of supervised learning, which consisted of matching pictures and descriptions, and this particular one offered to get to know how the visual features are the words in the captions. The paper will observe the process of generating English image captions with the CNN-LSTM models with the help of whose captions are the same as image captions, linguistically correct, and based on model inferences. The system is also able to support the top-k caption and provide feedback of the user to promote quality evaluation. The experimental findings demonstrate the fact that the method produces captions that are reasonable and consistent to contents of a large variety of images, to the benefit of the worth of CNN-LSTM encoder-decoder scheme of image captioning. Keywords - Image Captioning, Deep learning, CNN-LSTM, Encoder-Decoder Architecture, Visual Feature Extraction, Natural Language Generation, Multimodal Learning, Sequence Modeling