As we all know, population aging is a significant challenge faced by Chinese society, and ensuring the health and safety of elderly individuals has become an urgent topic of concern. Within the context of family safety, elderly individuals often experience falls or fainting due to age-related physical decline or underlying medical conditions. In response to this phenomenon, this paper presents a method for enhancing the wellbeing of family members by utilizing the YOLOv5 model to detect falls. Moreover, due to the built-in capability of YOLOv5 to read video from webcam, this technology can also be integrated into loT devices, turning these devices into a part of smart homes. Considering the specific nature of the home environment, CAU CAFall is considered to be the most suitable dataset. Various variations of the YOLOv5 model are experi-mented on a CAUCAFall dataset and achieve promising results. The YOLOv5x model achieved a precision of 82.2%, while the YOLOv5s model, with improved running speed, achieved an precision of 79.6%. Finally, we explored and selected the most suitable YOLOv5 model for home fall detection considering comprehensive evaluation metrics, and it is YOLOv5s.