US 11,704,355 B2
Method and system for producing story video
Junghwan Jin, Seongnam-si (KR); Sungwook Kim, Seongnam-si (KR); Sangho Choi, Seongnam-si (KR); Byung-Sun Park, Seongnam-si (KR); Wonhyo Yi, Seongnam-si (KR); Seongyeop Jeong, Seongnam-si (KR); Noah Hahm, Seongnam-si (KR); Jimin Kim, Seongnam-si (KR); and Hyeongbae Shin, Seongnam-si (KR)
Assigned to Snow Corporation, Gyeonggi-do (KR)
Filed by Snow Corporation, Seongnam-si (KR)
Filed on Sep. 20, 2019, as Appl. No. 16/577,759.
Application 16/577,759 is a continuation of application No. PCT/KR2017/003117, filed on Mar. 23, 2017.
Prior Publication US 2020/0019812 A1, Jan. 16, 2020
Int. Cl. G06F 16/438 (2019.01); G06N 20/00 (2019.01); G06V 20/40 (2022.01); G06V 20/00 (2022.01)
CPC G06F 16/4393 (2019.01) [G06N 20/00 (2019.01); G06V 20/35 (2022.01); G06V 20/41 (2022.01); G06V 20/47 (2022.01)] 10 Claims
OG exemplary drawing
 
1. A method of producing a story image, the method comprising:
collecting a plurality of images;
extracting caption information and meta information from each of the collected plurality of images;
classifying the collected plurality of images based on the extracted caption information and meta information, wherein the classifying of the collected plurality of images includes classifying each collected image into a category, from among plurality of categories;
determining a theme of a story to be produced based on the classified images, the theme being associated with one or more templates;
selecting a template from among the one or more templates associated with the determined theme of the story, each template from among the one or more templates including,
a plurality of frames for arranging images, and
for each frame from among the plurality of frames, at least one keyword or sentence corresponding to the frame;
for each frame from among the plurality of frames included in the selected template, selecting an image from among the collected plurality of images, based on the at least one keyword or sentence corresponding to the frame and at least one of the caption information and the meta information extracted from the selected image; and
generating the story image by arranging the selected images in the frames, from among the plurality of frames included in the selected template, for which the selected images were selected,
wherein, for each image from among the collected plurality of images,
the caption information extracted from the image includes text information acquired through a machine learning-based image analysis of the image,
wherein each category, from among the plurality of categories, corresponds to a theme, and
wherein the determining of the theme of the story includes,
determining a category, from among the plurality of categories, into which the most images are classified, from among the collected plurality of images,
determining, as the theme of the story, the theme corresponding to the determined category.