M-Stock: Automate Photo Assessment for Enhancing Student Photography Skills with Deep Learning

Presently, digital photography courses are being offered both in traditional university classrooms and through online platforms. Evaluating students' work in these courses, particularly when instructed by photography experts, presents challenges, especially in larger classrooms with numerous students. The difficulties include the inability to promptly assess and provide feedback due to the volume of work, leading to delays in meeting students' needs. Consequently, students are unable to receive immediate feedback on their submissions, causing a lag in their skill development. To address this issue, we propose the M-Stock system, an Automatic Photo Assessment for Enhancing Student Photography Skills with Deep Learning. This system operates as a web application and is compatible with all platforms. It utilizes Convolutional Neural Networks (CNN) for photo classification and evaluation. Through testing the classification efficiency, the system demonstrated an average accuracy of 97.18%, with an average classification speed of 46.1 milliseconds per image. This automation aims to streamline the evaluation process, ensuring timely feedback and fostering the timely development of students' abilities.