Human action recognition, storage and retrieval by combining edge and cloud computing to maintain privacy

Various embodiments of predicting human actions are disclosed. In one aspect, a human motion prediction system first receives a sequence of video images including at least a first person. Next, for each image in the sequence of video images, the human motion prediction system detects a first person in the video image; a skeletal graph of the detected first person is then extracted from the image of the detected first person, wherein the skeletal graph is composed of a set of human keypoints of the detected first person. Next, the human motion prediction system combines the detected skeletal graphic sequences of the first person extracted from the sequence of video images to form a first skeletal sequence of the detected first person depicting continuous motion of the detected first person. The human motion prediction system then transmits the detected first skeletal sequence of the first person to a server, wherein transmitting the detected first skeletal sequence of the first person to ...