Skeletal Hand Classification

Overview

A skeleton-based neural network that classifies hand poses live from a webcam. Instead of learning from pixels, it learns from the positions of the hand's joints, which keeps the model small and fast.

What I built

• Wrote a capture tool that records labeled hand-landmark data from a webcam.
• Trained Keras models on the landmark data to classify ASL letters.
• Built live software that draws the detected hand and shows the predicted label in real time.