Upright adjustment with graph convolutional networks
This addresses the upright adjustment problem for 360 images, which is an incremental improvement in computer vision.
The paper tackles the problem of upright adjustment for 360 images by proposing a method that combines a CNN and a GCN with a novel loss function for spherical distributions, and it outperforms fully connected-based methods in experiments.
We present a novel method for the upright adjustment of 360 images. Our network consists of two modules, which are a convolutional neural network (CNN) and a graph convolutional network (GCN). The input 360 images is processed with the CNN for visual feature extraction, and the extracted feature map is converted into a graph that finds a spherical representation of the input. We also introduce a novel loss function to address the issue of discrete probability distributions defined on the surface of a sphere. Experimental results demonstrate that our method outperforms fully connected based methods.