A Dual-Module System for Copyright-Free Image Recommendation and Infringement Detection in Educational Materials

Area of Science:

• Artificial Intelligence
• Educational Technology
• Copyright Law

Background:

• Images are vital for education but pose copyright risks.
• Unauthorized image use leads to legal challenges for educators.
• Existing solutions for image copyright verification are insufficient.

Purpose of the Study:

• To develop a dual-module AI system for educators to detect image copyright infringement.
• To provide a seamless solution for finding copyright-free alternative images.
• To enhance legal compliance in the creation of educational content.

Main Methods:

• A copyright infringement detection module using a Convolutional Variational Autoencoder (CVAE).
• An image retrieval module employing a Vision Transformer (ViT)-based hashing model.
• System evaluation using benchmark datasets (Flickr25k) and a user study with educators.

Main Results:

• The system achieved a mean Average Precision (mAP) of 0.812 on the Flickr25k dataset.
• A user study with 65 teachers showed high satisfaction with copyright detection and ease of use.
• The system effectively identifies copyright infringement and suggests suitable alternatives.

Conclusions:

• The developed dual-module system successfully addresses copyright infringement issues for educators.
• The AI-powered tool aids in creating legally compliant educational materials.
• The system offers a practical and user-friendly solution for copyright management in education.