Semantic class positioning digital environments

Abstract

Semantic segmentation techniques and systems are described that overcome the challenges of limited availability of training data to describe the potentially millions of tags that may be used to describe semantic classes in digital images. In one example, the techniques are configured to train neural networks to leverage different types of training datasets using sequential neural networks and useof vector representations to represent the different semantic classes.

Description

Background technique [0001] Semantic segmentation has achieved great progress with the advancement of neural networks in order to locate parts of digital images that correspond to semantic categories. For example, a computing device may use machine learning to train a neural network based on training digital images and labels identifying semantic categories exhibited by the digital images. Semantic categories can be used to identify specific objects included in the digital image, feelings evoked by the digital image, and the like. The model, once trained, is configured for use by a computing device to identify locations in the digital image that correspond to semantic categories. [0002] However, conventional techniques require labels and examples of associated digital images for each semantic category to be trained. Thus, traditional techniques are challenged by the limited availability of training data, which is further exacerbated by the multiple labels that can be used ...

AI Technical Summary

Problems solved by technology

For example, a traditional model trained by a computing device using machine learning for the semantic concept "person" may fail for the semantic concept "person" because the traditional model cannot recognize how related two semantic categories are to each other

Images