A Unified Framework for Multigrid Neural Network Architectures

Abstract

1. A method comprising: receiving an image as an input; providing a neural network structure including a plurality of multi-layer multi-scale neural networks, the plurality of multi-layer multi-scale neural networks being arranged in sequence by laterally connecting corresponding scale level layers between each two adjacent multi-layer multi-scale neural networks in a sequence of the plurality of multi-layer multi-scale neural networks; and, in a training phase, training the neural network structure on a training dataset to obtain a trained machine learning model configured to perform a computer vision task, the computer vision task comprising outputting at least one of: (i) classification of the image into one class of a set of two or more classes; (ii) segmentation of at least one object in the image; and (iii) detection of at least one object in the image.

Images