Using Convolutional Neural Networks in Robots with Limited Computational Resources: Detecting NAO Robots while Playing Soccer

摘要

The main goal of this paper is to analyze the general problem of usingConvolutional Neural Networks (CNNs) in robots with limited computationalcapabilities, and to propose general design guidelines for their use. Inaddition, two different CNN based NAO robot detectors that are able to run inreal-time while playing soccer are proposed. One of the detectors is based onthe XNOR-Net and the other on the SqueezeNet. Each detector is able to processa robot object-proposal in ~1ms, with an average number of 1.5 proposals perframe obtained by the upper camera of the NAO. The obtained detection rate is~97%.