A Transformer-Based Deep Learning Model for Arabic Image Captioning

Image captioning is a crucial process that connects text to images. Transformer-based deep learning models have successfully created captions, especially for Arabic. This paper introduces the Attention-Based Transformer Model for Arabic Image Captioning (ARTIC), which uses a deep learning convolutional neural network for feature extraction and a transformer encoder-decoder architecture for textual captions. The model uses an ensemble learning approach with a voting mechanism to select the caption with the highest bilingual evaluation understudy (BLEU) score. The model achieved the best BLEU-1, CIDEr, and ROUGE scores, while other metrics like BLEU-2 and METEOR achieved competitive rates.