Open
Description
Description
Currently, both MediaPipe and the Pointnet model are inference on the server side (see #6). We should look into using Tensorflow.js to classify the landmarks into English alphabet on the client side itself. This will significantly reduce the load on the server and make it accessible to more people around the world.
Alternatively, if Tensorflow.js is not the optimal approach, we can transmit landmarks to the server using websockets and classify them server-side, but this is not the ideal approach.