READ BEFORE BIDDING: Build machine learning to recognize hand & finger gestures in video

You need to have experience with machine learning and video. If not, please don't waste our time. Be prepared to talk in depth about your ML experience with video and the applications you have built. We need a system to train and recognize hand and finger gestures from a live video feed. The app needs to select a camera and watch the live video stream looking for specific gestures. The app needs to first train on specific gestures that can later be recognized When a gesture is recognized, a small video overlay should present itself at a pre-configured location. The overlay should represent the gesture detected. We can have an artist create these. For initial testing, text overlay of the gesture name is sufficient. The app should leverage a GPU as much as possible and should use generic code that is not tied to any manufacturer's GPU. We are open to using public domain code as long as it does not preclude building commercial applications. The end goal is to leverage the gesture recognition engine by other applications. The detected gesture will execute a macro that includes either a keyboard command or opens a URL To show that you have read this fully, put the name of you favorite or most used machine learning framework as the first words in you bid. Don't write "Hello", OR "my favorite is ____" etc. It should be the first words in your bid. This will prove you have attention to detail and did not spam bid. In this project we are looking for a proof of concept result. Get our attention. Talk about what tools you would use on this project. Generic bids will be ignored. Show us you are an expert. Thanks and good luck!