Published Paper Details:

Abstract: With the development of deep learning, the combination of computer vision and natural language process has attracted a lot of attention in the last few years. Caption embedding is representative of this file, which enables the computer to learn to use one or more sentences to understand the visual content of the image. The efficient process of producing high-definition semantics is required not only for the recognition of an object and the scene, but also for the ability to analyze the state, attributes and relationships between these objects. Although captioning is a difficult and difficult task, many researchers have found significant improvements. In this project, we describe in detail the three ways to insert image captions using deep neural networks CNN-RNN based, CNN-CNN-based framework and based on reinforcement. We then present the work representing the top three methods in a row, explain the metrics for testing and summarize the main advantages and disadvantages. Over the past few years, the problem of making automated descriptive sentences for photographs has gained a growing interest in natural language research and computer-based research. Image captioning is an important task that requires a basic understanding of images and the ability to produce descriptive sentences with appropriate and appropriate structure. In this study, the authors propose a hybrid system that uses the multilayer Convolutional Neural Network (CNN) to produce visual vocabulary and Long Short Term Memory (LSTM) to organize sound sentences using generated keywords. A convolutional neural network compares the image to a large database of training images, and creates an accurate description using trained captions.