Auricular deformities are one of the most common concerns in early childhood. These deformities can potentially lead to significant psychological distress, anxiety, and even hearing impairments in severe cases if left untreated. The non-surgical ear molding—a safe, effective and costless method requires to be done within the first 2–3 months of birth. However, the complexity associated with the auricle as whole and even within its sub-structures makes it challenging to accurately identify and classify them in newborns, particularly for less-experienced physicians. To address this critical need, an AI-aided approach for detecting and classifying auricular deformities in newborns is proposed, by making use of a recently released BabyEar4k Dataset that comprises of 3,852 high quality ear images acquired from 1,926 newborns. In this chapter, various single-stage and two-stage object detection models and their variants are fine-tuned and evaluated. The detection performance is assessed using the metrics: mAP50, mAP50-95 and classification performance using the metrics: accuracy, precision, recall and F1-score. The experiments resulted in YOLO-v8-L achieving the highest classification accuracy of 0.79 and Faster R-CNN with ResNet-50 backbone achieving precise localization with mAP-50 of 0.84. Additionally, D-RISE, an explainability technique is also incorporated, to generate visual explanations of model’s decision logics and highlight the regions that influenced the model’s decision, which is crucial especially in clinical settings.