AI-Based Photography Assessment System using Convolutional Neural Networ

Providing timely and meaningful feedback in photography education is challenging, particularly in large classes where manual assessment can delay skill development. This paper presents M-Stock, an AI-based automated photo evaluation system that uses Convolutional Neural Networks (CNNs) to assess student photography assignments on web browser. M-Stock evaluates both technical aspects (such as lighting, composition, and exposure) and creative elements, providing students with real-time, formative feedback. The system was trained on a diverse dataset, including student submissions and commercial standards, achieving an overall accuracy of 97.18% with an average prediction speed of 46.1 milliseconds per image. Experiments assessed the system’s performance across varying resolutions and batch sizes, confirming its scalability and suitability for real-time classroom use. Additionally, a pilot study with students indicated that M-Stock’s feedback positively impacted their technical skills and encouraged self-directed learning. The results demonstrate M-Stock’s potential as a transformative tool for photography education, combining high accuracy, immediate feedback, and pedagogical value to support continuous learning. Future improvements will focus on refining creative assessments and expanding the system’s applicability to other visual arts disciplines.