International Research Journal of Education and Technology

The rapid advancements in computer vision and natural language processing have enabled machines to interpret and describe visual content with remarkable accuracy. Image captioning, which involves generating meaningful textual descriptions for images, has emerged as a significant research area with wide-ranging applications in accessibility, surveillance, e- commerce, education, and digital media management. This paper presents an in-depth exploration of the design, development, and evaluation of an Image Caption Generator built using deep learning techniques. The proposed system integrates convolutional neural networks (CNNs) for visual feature extraction and recurrent neural networks (RNNs) with Long Short- Term Memory (LSTM) units for natural language generation. Datasets such as the Flickr Image Captioning Dataset and Microsoft COCO are used to train the model. Through the fusion of visual encoding and linguistic decoding, the system creates contextually relevant and grammatically coherent descriptions. Extensive experimentation demonstrates that the architecture performs effectively across diverse image categories. The paper concludes by discussing potential improvements using Transformer-based encoder–decoder structures and Vision-Language Models (VLMs), highlighting the future of multimodal AI research.

Keywords: Image captioning, deep learning, convolutional neural networks (CNN), recurrent neural networks (RNN), Long Short-Term Memory (LSTM), encoder–decoder architecture, vision–language models, natural language processing (NLP), computer vision, semantic image