Training data generating device, method, and program, and crowd state recognition device, method, and program
AssigneeNEC Corporation
InventorHiroo IKEDA
Life of this patent
Estimating the term…
Abstract
At least one storage stores a dictionary having information corresponding to crowd state images, and rectangular regions indicating portions of a given image to be recognized for crowd states, wherein a size of each of the rectangular regions is predetermined. At least one processor extracts regions, which correspond to the rectangular regions, from the given image, and recognizes crowd states in the extracted regions based on the dictionary.
Claims12
A crowd state recognition system comprising:
- at least one storage storing:
- a dictionary having information corresponding to crowd state images, and
- rectangular regions indicating portions of a given image to be recognized for crowd states, wherein a size of each of the rectangular regions is predetermined;
- and
- at least one processor configured to:
- extract regions, which correspond to the rectangular regions, from the given image, and
- recognize crowd states in the extracted regions based on the dictionary.
A crowd state recognition method, comprising:
- storing, in at least one storage,
- a dictionary having information corresponding to crowd state images, and
- rectangular regions indicating portions of a given image to be recognized for crowd states, wherein a size of each of the rectangular regions is predetermined;
- and
- extracting, by at least one processor, regions, which correspond to the rectangular regions, from the given image, and
- recognizing crowd states in the extracted regions based on the dictionary.
A non-transitory computer-readable recording medium in which a crowd state recognition program is recorded, the crowd state recognition program causing a computer, which comprises at least one storage storing a dictionary having information corresponding to crowd state images, and rectangular regions indicating portions of a given image to be recognized for crowd states, wherein a size of each of the rectangular regions is predetermined, to perform:
- extracting regions, which correspond to the rectangular regions, from the given image, and
- recognizing crowd states in the extracted regions based on the dictionary.
Dependencies are inferred from the claim text.