The assessment of creativity in early childhood is important for identifying developmental potential, yet existing methods such as manual scoring of the Test for Creative Thinking–Drawing Production (TCT-DP) are time-consuming and susceptible to subjectivity. This study proposes an automated framework combining Discrete Wavelet Transform (DWT) and Convolutional Neural Networks (CNNs) to support creativity assessment in children aged 5–8 years based on TCT-DP drawings. A dataset of 100 drawings, scored by expert raters, was preprocessed and decomposed using a two-level Daubechies db4 wavelet to extract spatial-frequency features. These features were used as inputs to a CNN model trained to classify creativity levels. Model performance was evaluated using accuracy, F1-score, ROC-AUC, and Pearson’s correlation with expert scores. The proposed model achieved 87% accuracy and a correlation of r = 0.74, indicating moderate agreement with expert ratings. While results suggest potential for improving efficiency and consistency, findings remain exploratory due to limited sample size.