Classification of Objects Using Neuromorphic Camera and Convolutional Neural Networks

The classification of objects is a field very well explored by Computer Vision and has achieved excellent results over the last decade. The sophistication of biological systems has led us to the development of bioinspired technologies through Neuromorphic Engineering that proposes to develop robotic systems with operation inspired by the physiological processes found in nature. Seeking to combine the high speed of information processing, with the low dimensionality of the data and a reduced computational cost, we combine a neuromorphic vision sensor (DVS128) with a Convolutional Neural Network (CNN) to classify images of nine different objects. Our deep learning model achieved 75.31% accuracy when performing network validation using the holdout method.