Smart Baby Monitoring System using YOLO $\mathrm{V}7$ Algorithm

Monitoring a newborn live and in real-time is the most important step to do in order to know their status while we are away. Because the camera's angle is fixed and the baby's position moves, they might not be included in routine baby monitoring. In this research, a novel approach based on IoT and deep learning is offered as a solution to this problem. The YOLO model and the IOT model have both been used in the proposed work to recognize the positioning of the baby and to automatically modify the camera's movement in that direction. Images of babies in various environments are gathered from Google and Kaggle. The Roboflow framework is used to label the collected photos. YOLOv4 -tiny and YOLOv7 are used to implement the suggested approach. YOLOv7 is an incredibly quick object detector that is more accurate and appropriate for embedded devices. Librosa is used for audio training and categorization. Additionally, in instances involving real-time video processing, each model makes quick predictions. The outcome of the proposed system is examined with the mean Average Precision (mAP). With a higher mAP score, the model is thought to be more capable of detection. With us YOLOv4-tiny model generated a mAP value of 36.02% with an inference time of 17 milliseconds, and our YOLOv7 model generated a mAP value of 67.4% with such an inference time of 2.8 milliseconds.