Indoor scene semantic annotation method

Abstract

The indoor scene semantic annotation method comprises the following steps: inputting a color picture and a depth picture; entering the neural network, wherein the color picture and the depth picture pass through a conv1 and a conv2 _ x respectively; entering a first attention mechanism module ARF _ 1, and obtaining a feature map through calculation of the ARF _ 1; entering conv3 _ x to carry out convolution calculation; entering a second attention mechanism module ARF _ 2, and obtaining a feature map through calculation of the ARF _ 2; entering conv4 _ x to carry out hole convolution calculation; entering a third attention mechanism module ARF _ 3, and obtaining a feature map through calculation of the ARF _ 3; entering conv5 _ x to carry out hole convolution calculation; entering an attention mechanism mixing module ARMF for calculation; entering a space pyramid module SPP to realize multi-level context information fusion; obtaining a semantic annotation result graph.

Description

technical field [0001] The invention relates to the technical fields of multimedia technology and computer graphics, in particular to a semantic labeling method for indoor scenes. Background technique [0002] Scene labeling, or scene parsing, is to label each pixel in the image with the object category label it belongs to. Scene semantic annotation is a challenging task since it combines the traditional problems of detection, segmentation, and multi-label recognition in a single process. High-quality scene annotation is beneficial for intelligent tasks such as robot task planning, pose estimation, plane segmentation, context-based image retrieval, and automatic photo adjustment. [0003] The previous scene identification work can be divided into two categories: indoor scene and outdoor scene according to the target scene. Compared with outdoor scene labeling, indoor scene labeling is more challenging because of the larger set of indoor scene semantic labels, more serious ...

AI Technical Summary

Problems solved by technology

However, these methods decouple context modeling from convolutional feature learning, which may pose the problem of finding suboptimal solutions due to small differences in feature representations.

Another type of method uses a cascaded recurrent neural network with a gate structure, such as a long short-term memory (LSTM) network, to strengthen context modeling, but the method of building a context model based on a recurrent neural network only fuses context information at a specific level of the network. , did not increase the receptive field in the feature learning process, nor did it improve the feature expression in the feature learning process

Images