What type of study is this?

This is a Literature Review study.

September 17, 2025Open Access

A Review on Deep Learning based Image Caption Generation Methods

Puntos clave

Recent advancements in image captioning techniques utilize deep learning, improving caption generation accuracy.
State-of-the-art models leverage attention mechanisms, with evaluation based on datasets like MS-COCO and Flickr8k.
The proposed conceptual model integrates cnn for feature extraction with transformers for more fluent captions.
Critical evaluation addresses the strengths and weaknesses of various methodologies, guiding future research directions.

Resumen

Abstract - This review investigates recent advancements in Deep Learning based image caption generation methodologies. The progression from early encoder-decoder models to attention-based systems and transformer frameworks is examined, alongside notable datasets and evaluation metrics. Strengths and shortcomings of leading methods are analyzed to provide researchers with a consolidated understanding of current trends. The proposed conceptual model integrates Convolutional Neural Networks for visual feature extraction with transformer mechanisms for sequence generation, aiming to produce accurate, fluent, and context-aware captions. Comparative discussions of state-of-the-art techniques are supported with evaluations based on benchmark datasets like MS- COCO and Flickr8k, emphasizing attention strategies including soft, hard, and adaptive variations. Key Words: Image Captioning, CNN, Transformer, Attention, Deep Learning.

Leer artículo completoexternamente

Preguntar a la IA

Me gusta

Guardar

Ver artículo completo