AI-driven video summarization for optimizing content retrieval and management through deep learning techniques

Area of Science:

• Artificial Intelligence
• Computer Science
• Information Science

Background:

• Growing volume and complexity of digital video content challenge traditional archive management.
• Existing methods for video organization and retrieval often require significant human input or use simplistic algorithms.
• Need for scalable and efficient solutions to manage large, heterogeneous media archives.

Purpose of the Study:

• To develop a novel artificial intelligence (AI) framework for enhanced organization and retrieval of video content.
• To address the limitations of current methods in handling large-scale video archives.
• To explore how AI can support sustainable digital practices in media management.

Main Methods:

• Utilized convolutional neural networks (CNNs) and long short-term memory (LSTM) networks for extracting frame-level and temporal video features.
• Integrated Residual Networks 50 (ResNet50) for improved content representation.
• Employed two-frame video flow to enhance system performance.

Main Results:

• Achieved high performance metrics: 79.2% precision, 86.5% recall, and 83% F-score.
• Validated the framework on diverse datasets including YouTube, EPFL, and TVSum.
• Demonstrated the system's capability to minimize data duplication and optimize resource usage.

Conclusions:

• The proposed AI framework offers a scalable and effective solution for managing large media collections.
• The system contributes to promoting sustainable digital practices through optimized resource utilization.
• AI advancements can significantly enhance organizational environmental, social, and governance (ESG) outcomes via improved content management.