Phantom image classification

摘要

A computing device uses image metadata to perform a first sub-classification of an image, wherein the first sub-classification is deterministic, then uses a combination of image data and the image metadata to perform a second sub-classification, further in a third sub-classification, uses image data to make a probabilistic classification of an image according to a likely phantom type. Alternatively or additionally, a computing device receives a set of image analysis data, identifies an image analysis model to be trained, using the image analysis data, determines that the image analysis model can include an image permutation, and trains a new image analysis model including the image permutation.

主权项

1. A system, comprising a computing device that includes a processor and a memory, the memory storing instructions executable by the processor, the instructions including instructions for:
using image metadata to perform a first sub-classification of an image, wherein the first sub-classification is deterministic; based in part on the deterministic first sub-classification, using a combination of image data of the image and the image metadata to perform a second sub-classification, and in a third sub-classification, based in part on the second sub-classification, using the image data to make a probabilistic classification of the image according to a likely phantom type.