A gesture classifier created using OpenCV and a basic CNN.
This was made as a demonstration for my work with the Gesture Detection Project Team at UTMIST. It is a basic CNN trained on a dataset of grayscale images, with the trained model then being used to predict the gestures from a webcam input. It's not perfect, due to the limitations of OpenCV's thresholding capabilities and some discrepancies within the dataset, but can easily be adapted for colour images. The goal is to eventually adapt this for a trained 3D CNN model.
Dataset can be found here: https://www.kaggle.com/gti-upm/leapgestrecog