Skip to main content Skip to navigation

ArtMentor: AI-Assisted Evaluation of Artworks to Explore Multimodal Large Language Models Capabilities

Project Overview

The document explores the application of Multimodal Large Language Models (MLLMs), particularly through the ArtMentor system, in the field of art education. It emphasizes the role of generative AI in facilitating art evaluation by assisting teachers in recognizing elements within artworks, generating insightful reviews, and proposing constructive feedback. The ArtMentor system not only enhances the accuracy and efficiency of art assessments but also fosters a collaborative interaction between educators and AI. By leveraging user data, the system provides tailored suggestions for improvement, which has the potential to significantly boost students' creativity and artistic skills. Overall, the integration of generative AI in art education represents a transformative approach to teaching, enabling a more interactive and supportive learning environment.

Key Applications

ArtMentor System

Context: Art education for teachers, art mentors, and school administrators, focusing on art evaluation and feedback for elementary school students.

Implementation: The ArtMentor system integrates machine learning language models (MLLMs) to assist in evaluating artworks through entity recognition. It generates prompts based on user-provided data and delivers personalized feedback and suggestions for artistic development and assessment.

Outcomes: ['Improved efficiency in art evaluation', 'Enhanced understanding of artistic elements', 'Improved feedback mechanisms for students', 'Enhanced creativity', 'Personalized suggestions for artistic development']

Challenges: ['Challenges in accurately recognizing broader entities', 'Potential for over-granulation in evaluations', 'Potential inaccuracies in entity recognition', 'Need for further improvements in usability']

Implementation Barriers

Technical/Technological Barrier

MLLMs sometimes overlook broader artistic entities and may over-focus on local features. Additionally, inaccuracies in entity recognition within artworks may affect the quality of feedback provided by the system.

Proposed Solutions: Further optimization of models to better differentiate between complex art styles and reduce over-granulation in entity recognition, along with continuous optimization of the system based on user feedback and performance evaluations.

Methodological Barrier

The process-oriented data in art evaluation is complex and difficult to analyze.

Proposed Solutions: Developing innovative approaches in HCI to enhance usability and effectiveness of process mining tools in educational environments.

Usability Barrier

Potential difficulties users may face in navigating the system and effectively utilizing its features.

Proposed Solutions: Gathering user input for interface improvements and user experience enhancements.

Project Team

Chanjin Zheng

Researcher

Zengyi Yu

Researcher

Yilin Jiang

Researcher

Mingzi Zhang

Researcher

Xunuo Lu

Researcher

Jing Jin

Researcher

Liteng Gao

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Chanjin Zheng, Zengyi Yu, Yilin Jiang, Mingzi Zhang, Xunuo Lu, Jing Jin, Liteng Gao

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies