Practical Techniques for Using Neural Networks to Estimate State From Images
This paper is a part of our work on using neural networks to train robots to perform problems. It includes practical techniques for dealing with huge images, chaotic systems, and more.
Abstract: An important task for training a robot (virtual or real) is to estimate state. State includes the state of the robot and its environment. Images from digital cameras are commonly used to monitor the robot due to the rich information, and low-cost hardware. Neural networks excel at categorizing images, and should prove powerful to estimate the state of the robot from these images. There are many problems that occur when attempting to estimate state with neural networks, including high resolution of images, training time, vanishing gradient, and more. This paper presents several practical techniques for facilitating state estimation from images with neural networks.