Exponential Sailfish Optimizer-based generative adversarial network for image annotation on natural scene images.

Area of Science:

• Computer Science
• Artificial Intelligence
• Image Processing

Background:

• Automatic image annotation is crucial for content-based image retrieval.
• Existing methods often overlook positional and descriptive information.
• Generative Adversarial Networks (GANs) show promise but require optimization.

Purpose of the Study:

• To develop an efficient and accurate automatic image annotation technique.
• To integrate novel optimization strategies with GANs for improved performance.
• To leverage additional image information beyond semantic and visual content.

Main Methods:

• Implementation of Exponential Sailfish Optimizer (ESFO) for training GAN classifiers.
• Combination of Exponentially Weighted Moving Average (EWMA) and Sailfish Optimizer (SFO) within ESFO.
• Utilizing Grabcut for foreground and background image extraction.
• Employing DeepJoint segmentation for image partitioning.

Main Results:

• The ESFO-based GAN achieved high performance metrics on the flicker dataset.
• Maximum F-Measure reached 98.37%, precision 97.02%, and recall 96.64%.
• The method effectively utilized positional and descriptive data for annotation.

Conclusions:

• The developed ESFO-based GAN provides a robust solution for automatic image annotation.
• The integration of ESFO significantly enhances GAN performance in image recognition tasks.
• This approach offers a valuable advancement for large-scale image dataset management and retrieval.