A Study on CNN Model with Multiple Inputs for Detection of Driving Area

In this paper, we propose a method to distinguish the drivable and non-drivable areas of a mobile robot by learning road driving image data of an autonomous vehicle. Through recognition of the drivable area of vehicles and mobile robots, it is intended to improve performance more stably when driving and mapping. We propose a CNN model for segmenting the driving area by learning road driving image data. The proposed model proceeds with a performance evaluation of the model through one input and multiple inputs. We shows that the proposed model preserves spatial information and reduces the network parameters.