Enhancing Image Caption Generation Through Context-Aware Attention Mechanism
Published in Heliyon, 2024
This research article proposes a context-aware attention mechanism for image caption generation. Standard attention-based captioners attend to image regions independently of the language context produced so far; the proposed mechanism conditions visual attention on the previously generated tokens, encouraging captions that are more semantically grounded in the image and less prone to context drift or hallucinated content.
Published as a Research Article in Heliyon, Volume 10, Issue 17 — September 15, 2024.
Recommended citation: Bhuiyan, A. (2024). "Enhancing Image Caption Generation Through Context-Aware Attention Mechanism." Heliyon, 10(17). Research Article.