Hierarchical Interlinked Multi-scale Convolutional Network for Image Parsing

摘要

A disclosed facial recognition system (and method) includes face parsing. In one approach, the face parsing is based on hierarchical interlinked multiscale convolutional neural network (HIM) to identify locations and/or footprints of components of a face image, The HIM generates multiple levels of image patches from different resolution images of the face image, where image patches for different levels have different resolutions. Moreover, the HIM integrates the image patches for different levels to generate interlinked image patches for different levels, where interlinked image patches for different levels have different resolutions. Furthermore, the HIM combines the interlinked image patches to identify refined locations and/or footprints of components.

主权项

1. A system for parsing an image into components, the system comprising:
a hierarchical interlinked multiscale convolutional neural network (HIM) for locating the components from the image, the HIM comprising:
a level generator configured to receive the image and to generate N levels of image patches from the image, N>2, wherein the image patches for different levels n have different resolutions R(n) and the image patches for a level n are generated from the image resampled to resolution R(n);an interlinked combiner configured to receive the N levels of image patches from the level generator and to generate M levels of interlinked image patches from the N levels of image patches, 2