Computer Vision high capacity system needed

1- Ability to accept any video stream (live or stored) 2- Ability to mark any object in video to be tracked in the video, mark it by labeling, record position of object in the video. e.g. Video frame 1min 30sec. Object "abc" position "x,y" in that frame. Store this in the db for playback so next time no need to run the job again in CV. 3- Code should be written in a way that it can process millions of videos and identify\track millions of objects at a time.