首页>
外国专利>
Semantic scene segmentation using random multinominal logit (RML)
Semantic scene segmentation using random multinominal logit (RML)
展开▼
机译:使用随机多语言Logit(RML)的语义场景分割
展开▼
页面导航
摘要
著录项
相似文献
摘要
A computer implemented method for learning a Random Multinomial Logit (RML) classifier (126) for scene segmentation, the method comprising: receiving (210) an image training set, wherein the image training set comprises a plurality of digital representations of images (302), and an object of an image ( 302) of the image training set has a semantic identifier; generating (212) a plurality of Texton images (306) corresponding to the images (302) in the image training set, wherein a Texton image (306) of its corresponding image (302) in the image training set Image of pixels, and wherein each pixel value in a Texton image (306) is replaced with a representation of the pixel value of its corresponding image (302) in the image training set; choosing (214) one or more texture layout features (308) of the plurality of Texton images (306), wherein selecting one or more texture layout features (308) comprises: swapping a feature selected by the RML-K classifier (126) is currently in use against a randomly selected new feature based on the statistical significance of the currently used feature; learning (216) multiple multinomial logistic regression models of the RML classifier (126) based on the selected texture layout features; and evaluating (218) the performance of the plurality of multinomial logistic regression models based on the semantic identifiers of the objects in the image training set.
展开▼