SphereNet-pytorch
SphereNet-pytorch copied to clipboard
How to create an omnidirectional MNIST dataset
“For the classification task, we create an omnidirectional MNIST dataset (Omni-MNIST), where MNIST digits are placed on tangent planes of the image sphere and an equirectangular image of the scene is rendered at a resolution of 60 × 60 pixels.” I don't understand it, and hope to get a further explanation. Thank you very much!