Whether to use one image or multiple images when inferring MapNet

[cccccv-cm]

Hello, I have encountered a problem. When I reasoned about MapNet, I found that according to the settings in the configuration file, the input is a tensor of 3 pictures, but what is written in the paper is that the input of MapNet during reasoning is a picture, so what is going on and what should I do? ? Are the network inputs of PoseNet and MapNet the same or variable?