The Artificial Neural Network (ANN) is a computing system inspired by the human brain that uses interconnected nodes to recognize patterns and process data. It is the simplest component of various deep learning structures, featuring only a linear layer of neurons to calculate and make predictions.
While impressive on its own for more general-purpose tasks, the ANN does struggle from performing multi-dimensional data parsing and feature extraction especially for colored images. As shown in the tables below, the gesture and object identification tests where it barely scored correct results from varied objects it attempts to identify shows the struggles in doing so. This app will still include this model as an appropriate baseline to compare with the CNN.
The Convolutional Neural Netword (CNN) is a specialized type of deep learning algorithm designed to process grid-like data, such as images and videos. Due to its layers using matrix-based calculations to focus and extract features from both one-dimensional and multi-dimensional dataset, they are used mainly for tasks such as computer vision, image and audio processing.
The CNN shows better overall identification and processing with most of the letters involved. Compared to the ANN, it shows better performance in identifying overall gestures, although it still shows confidence drops when identifying similar-shaped gestures such as A and S. The CNN identification process is also challenged by the properties of the submitted image, such as other objects in the image, the distance of the hand from the camera (closer is better), and the brightness and noise involved in the image.
