Construction of a Smart Evaluation System for After-School Art Education Based on Educational Information Systems

Abstract

Over the past decade, data and computational power have exploded, advancing this field to the point where deep learning can have a significant transformative impact across many subfields of artificial intelligence (Liu et al., 2023a). Real-world applications now utilize Deep Neural Networks (DNNs) as a key to their innovations in image recognition, emotion detection, and intelligent systems (Liu et al., 2023b; Nabil et al., 2021). However, even though DNNs can produce impressive results, hidden flaws of DNN models can still generate incorrect outputs, leading to significant real-world consequences (Liu et al., 2022; Mo et al., 2018). As with ordinary software systems, there is a need for systematic testing of DNNs in order to increase reliability. However, unlike conventional code testing, the work typically lacks obvious expected outputs and is often expensive to test due to the need for manual coding (Guo et al., 2022; Guo et al., 2023; Li et al., 2022).