Todly: Multimodal Approach for Toddler Safety and Well-Being Using Advanced Machine Learning

This study presents “Todly,” an intelligent toddler monitoring system designed as a mobile application to ensure the safety and well-being of toddlers through cutting-edge technology. “Todly” integrates advanced machine learning algorithms and neural networks to enhance child safety and provide caregiver support. This approach features face emotion and gesture detection, danger object detection, crying pattern analysis, and a chatbot assistant for new mothers. These components work together seamlessly to offer caregivers real-time insights into a child's emotional state, identify potential hazards, understand the reasons behind a toddler's crying, and provide timely support and guidance. The architecture is designed to be scalable and adaptable, making it suitable for various environments, such as homes, daycare centers, and pediatric healthcare facilities. Its user-friendly interface ensures accessibility for users with different levels of technical expertise, while the robust support infrastructure facilitates smooth operation and maintenance. The integrated solution demonstrated strong performance across its various components, achieving an accuracy of 89.9% in object detection, 85% in emotion detection, and 90.8% in cry analysis, encompassing all cry types. The chatbot's mT5 model also achieved BLEU scores of 38.7 for Sinhala-to-English translation and 31.2 for English-to-Sinhala translation.