Image captioning is a crucial task at the intersection of computer vision and natural language processing (NLP), aimed at generating natural language descriptions for given images. This paper proposes a deep learning approach to image captioning by combining a pre-trained DenseNet for feature extraction with Long Short-Term Memory (LSTM) networks for generating descriptive captions. In the proposed method, DenseNet serves as a convolutional feature extractor, capturing rich and intricate visual features from images. These extracted features are then fed into an LSTM-based model, which generates captions by predicting one word at a time. The model follows a sequence-to-sequence learning architecture, where the input image features, and a sequence of words form the foundation for caption generation. To efficiently handle the dataset, we utilize a custom data generator that optimizes memory usage. The model is trained on a dataset of images paired with corresponding captions and evaluated using various performance metrics such as loss and accuracy. This study demonstrates that integrating dense convolutional networks with recurrent neural networks significantly enhances the quality of generated captions, providing valuable insights for applications in accessibility, image retrieval, and content generation.