Automated AI-Based Image Captioning: A Transformer-Based Approach for Natural Language Generation from Visual Data

Image captioning is essential in computer vision and Natural Language Processing (NLP) to produce relevant textual descriptions from visual information. This paper demonstrates a transformer architecture for deep-learning image captioning that uses attention mechanisms in Transformers to produce improved captions. The parallel processing capability of transformers makes them different from conventional CNN-RNN-based models and allows faster training with better contextual understanding. The research explores Vision Transformer (ViT) and Contrastive Language-Image Pretraining (CLIP) as transformer-based models that work with language models to create superior captioning results. The proposed methodology performs better than conventional models after demonstrating high results on benchmark datasets MS COCO and Flickr8k.Experimental evaluations demonstrate that our technique leads to enhanced scores for BLEU, METEOR and CIDEr metrics, thus proving its effectiveness.The paper investigates forthcoming prospects and present obstacles in automated image captioning technology.