Authors:

Lamiya Sajidbhai Laxmidhar

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

This project investigates the feasibility of emotion recognition in children's artwork using classical machine learning methods and hand- crafted features. Children's drawings are a widely studied medium of emotional expression in developmental psychology, yet they remain underexplored in computational emotion classification research. This project introduces CrayonClassifier, an end-to-end image classification pipeline designed to identify five core emotional categories-happiness, sadness, calmness, fear, and anger-from crayon-style children's drawings. The dataset was curated through targeted web search and organized into labeled emotional categories. To ensure robustness and class balance, data augmentation techniques such as geometric transformations were applied. Preprocessing was performed using OpenCV, with facial or visual region-of-interest (ROI) detection where applicable. Wavelet transforms were employed for feature extraction to capture both texture and spatial frequency information. Three classical machine learning models-Support Vector Machines (SVM), Logistic Regression, and Random Forest-were trained and evaluated. Hyperparameter tuning was conducted using GridSearchCV, and model performance was assessed via accuracy, precision, recall, and confusion matrix analysis. The finalized model was deployed through a Flask server and integrated with a web-based user interface developed using HTML, CSS, and JavaScript. CrayonClassifier demonstrates that with appropriate feature engineering and traditional ML models, it is possible to extract and classify affective signals from children's visual expression. The work has implications for interdisciplinary applications across child development research, emotional diagnostics, and human-centered AI. † Presenting Undergrad Author; ‡ Contributing Undergrad Author; * Undergrad Acknowledgment Keywords: Emotion Recognition; Image Classification; Children; Machine Learning; Feature Engineering

Source:

Purdue University / 2025

Topics:

No topics listed

Co-authors:

Lamiya Sajidbhai Laxmidhar