首页>
外国专利>
MODELING SEMANTIC CONCEPTS IN AN EMBEDDING SPACE AS DISTRIBUTIONS
MODELING SEMANTIC CONCEPTS IN AN EMBEDDING SPACE AS DISTRIBUTIONS
展开▼
机译:嵌入空间中的语义概念作为分布的建模
展开▼
页面导航
摘要
著录项
相似文献
摘要
#$%^&*AU2016259337A120170803.pdf#####Abstract Modeling semantic concepts in an embedding space as distributions is described. In the embedding space, both images and text labels are represented. The text labels describe semantic concepts that are exhibited in image content. In the embedding space, the semantic concepts described by the text labels are modeled as distributions. By using distributions, each semantic concept is modeled in the embedding space as a continuous cluster which can overlap other clusters that model other semantic concepts. Once an embedding space is trained, the embedding space can be used to discover text labels to describe content of an image. To use a trained embedding space to discover text labels that describe content of an image, multiple semantically meaningful regions of the image can be determined and corresponding text labels can be discovered in the trained embedding space for each of the regions.Inventor: Lin et al. Title: Modeling Semantic Concepts in an Embedding Space as Distributions 500 502 Process a training image having multiple text labels to generate a set of image regions that correspond to the respective multiple text labels 504 Embed the set of regions within an embedding space that is configured to embed both text labels and image regions mapped to the text labels 506 Apply label discovery techniques to a query image to map image regions of the query image to the text labels in the embedding space to discover text labels that correspond to the image regions 508 Annotate the query image with the discovered text labels to describe the content of the query image 510 Present the regions of the query image that correspond to the discovered text labels
展开▼