A Cross-Media Advertising Design and Communication Model Based on Feature Subspace Learning

Area of Science:

• Artificial Intelligence
• Computer Vision
• Machine Learning

Background:

• Traditional feature subspace learning models struggle to preserve local structure and discriminative properties of data after projection.
• Cross-media retrieval is crucial for modern advertising, but faces challenges in accurately correlating and retrieving information across different modalities like images and audio.

Purpose of the Study:

• To develop a discriminative feature subspace learning model for enhanced cross-media retrieval in advertising design.
• To improve the accuracy and efficiency of retrieving information between different media types, specifically images and audio.
• To advance the theoretical and practical application of digital media technology in commercial advertising.

Main Methods:

• Proposed a discriminative feature subspace learning model based on Low-Rank Representation (LRR) to maintain local sample structure and nearest-neighbor relationships.
• Utilized extreme learning machines and deep convolutional generative adversarial networks to improve cross-modal retrieval accuracy and explore correlations between different data modalities.
• Implemented an optimization algorithm based on similarity transfer for correcting clustering quality and active learning strategies to enhance retrieval efficiency with limited feedback.

Main Results:

• The proposed model accurately measures cross-media relevance, enabling effective mutual retrieval between image and audio data.
• Demonstrated improved cross-modal retrieval accuracy by mining deeper data features and maximizing correlations between modalities.
• Showcased enhanced efficiency in cross-media retrieval, particularly in scenarios with limited annotated samples.

Conclusions:

• The developed feature subspace learning model offers a significant advancement for cross-media advertising design and communication.
• The findings provide practical guidance for designers and artists in leveraging digital media technology for artistic and commercial applications.
• This research positively impacts commercial advertising by improving the utilization of digital media for design and communication.