A Software Retina for Egocentric & Robotic Vision Applications on Mobile Platforms
Jan Paul Siebert, Adam Schmidt, Gerardo Aragón-Camarasa, Nick Hockings, Xiaomeng Wang, W. Paul Cockshott · ENLIGHTEN (Jurnal Bimbingan dan Konseling Islam) · 2016
We present work in progress to develop a low-cost highly integrated camera sensor for egocentric and robotic vision. Our underlying approach is to address current limitations to image analysis by Deep Convolutional Neural Networks, such as the requirement to learn simple scale and rotation transformations, which contribute to the large computational demands for training and opaqueness of the learned structure, by applying structural constraints based on known properties of the human visual system. We propose to apply a version of the retino-cortical transform to reduce the dimensionality of the input image space by a factor of ex100, and map this spatially to transform rotations and scale changes into spatial shifts. By reducing the input image size accordingly, and therefore learning requirements, we aim to develop compact and lightweight egocentric and robot vision sensor using a smartphone as the target platform