3 Image caption generating system using convolutional neural networks and long short-term memory networks
P. Umamaheswari, D Vaishnavi · 2023
Human beings have the ability of easily determining the environment and the situations happening around them. Making a computer to automatically understand the surroundings can be done by training it with the images and ensuring that it describes the content of images. In this way, a computer will be able to learn the surroundings. Image caption generation is a process of creating appropriate captions for an image which is linguistically suitable and semantically precise for a scene and is also same to the understanding of humans. It is gaining attention because of its wide applications in scene recognition systems, human-robot communication systems, information retrieval systems and assistance systems for visually disabled persons. In this research, an image caption generating system is implemented using state-of-theart CNN model InceptionResnetV2 for image feature extraction (transfer learning), and long short-term memory networks for image caption generation. The model is trained using Flickr8k dataset and it is evaluated using metrics like BLEU and METEOR scores.