A scene classification algorithm of visual robot based on Tiny Yolo v2
Qiushi Pan, Yutong Guo, Zhiliang Wang · 2019
When robots and unmanned vehicles are running, they will perform corresponding actions according to different scenes. Traditionally, convolutional neural network is used to classify scene images with low accuracy. This paper proposes a kind of scene classification algorithm (end-to-end multi-object recognition scene classification algorithm, hereinafter referred to as EMORSC algorithm) based on the object recognition algorithm. It is to analyze the main object on the picture and classify it according to the scene in which the object is located. The final accuracy can be more than 80% higher than the classification result of convolutional neural network. In the future, this algorithm can be applied to intelligent logistics, intelligent driving and other fields, and has high practicability in these fields.