Indoor 360 panoramas have two essential properties.(1) The panoramas are continuous
and seamless in the horizontal direction.(2) Gravity plays an important role in indoor
environment design. By leveraging these properties, we present PanelNet, a framework that
understands indoor environments using a novel panel representation of 360 images. We
represent an equirectangular projection (ERP) as consecutive vertical panels with
corresponding 3D panel geometry. To reduce the negative impact of panoramic distortion …

S. To make the network aware of ERP distortion, we generate the geometric features in the
same size as the output of layer1 and add them together as the input of layer2. Note that the
output feature map of layer1 is in shapefc∈ RCc× He 4× I 4× N. For a ResNet-34 encoder,
Cc= 64. Thus, we generate the corresponding local and global 3D Cartesian coordinates of
each point from an image in resolution He