Enhanced ASD detection using 3D facial landmark localization with convolutional shape appearance model and graph-randomized XGBoost

Key Points

• Detection accuracy significantly improved through 3D facial landmark localization and graph-randomized xgboost.
• The proposed method achieved an accuracy rate of 95.6% with a dataset of 250 images.
• This analysis utilizes a convolutional shape appearance model for accurate facial feature extraction.
• The findings highlight the potential for more efficient ASD screening and early intervention.