This project aims to achieve highly accurate hand pose recognition with deep learning, particularly on several specific hand gestures. The project will be divided into three stages.
In the first stage, our goal is to build a deep learning model that can detect or segment human hands from the input picture. Accuracy is of great importance to this stage since it is the base of all following stages.
In the second stage, we intend to build a model that takes in the output of stage 1 and mark down the joints and fingertips. We expect that our model can achieve better performance than the state-of-the-art in terms of time and accuracy.
Our final objective is to develop an application that can timely recognize and respond to certain specific hand gestures and carry out the requested tasks. For example, one of the hand gestures that we target at is the “thumbs-up” gesture, which can be recognized by our model and contribute one like to our final year project page. The images of the hands should be able to be captured by an ordinary camera that is available on modern smart devices, and the analyzed results can be used to enhance user experience in interacting with these smart devices.