SkyEye is a deep learning based system, which includes a self-made sensor "AI Image Sensor for human falling", a cloud server for storing and posting fall event information, and a mobile APP for communicating with users. First, the AI sensor captures the live video through the image module, then identify whether the elder have fallen in the embedded system. Two CNN layers included in deep learning image recognition model. The PAFs of the first CNN layer are responsible for capturing the position and direction of the human body 2D vector. The CMP of the second CNN layer generates a heat map of a specific part. Finally, combining these two results can be used to construct the human skeleton. Then, determine whether the elder fall with the fall detection algorithm. If so, the embedded system transmits the signal of the fall event to the cloud server via the PoE network. Finally, pushes the event to mobile app via 4G network instantly.