US 12,363,399 B2
Video clipping method and model training method
Chengdong Wei, Beijing (CN); Jiechao Ding, Beijing (CN); and Rui Gao, Beijing (CN)
Assigned to Beijing Baidu Netcom Science Technology Co., Ltd., Beijing (CN)
Filed by Beijing Baidu Netcom Science Technology Co., Ltd., Beijing (CN)
Filed on Jan. 3, 2023, as Appl. No. 18/092,608.
Claims priority of application No. 202210023573.0 (CN), filed on Jan. 10, 2022.
Prior Publication US 2023/0224556 A1, Jul. 13, 2023
Int. Cl. H04N 21/845 (2011.01); G06V 30/148 (2022.01); G06V 40/20 (2022.01); H04N 21/466 (2011.01)
CPC H04N 21/8456 (2013.01) [G06V 30/153 (2022.01); G06V 40/20 (2022.01); H04N 21/4666 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A video clipping method, comprising:
acquiring, by a video clipping device, interaction behavior data for an original video file;
determining, by the video clipping device, interaction heat at respective time points of the original video file, according to the interaction behavior data;
selecting, by the video clipping device, N time points with highest interaction heat, to take the selected time points as interest points of the original video file, wherein N is a positive integer;
clipping, by the video clipping device, the original video file based on the respective interest points, to obtain N clipped video files;
performing, by the video clipping device, secondary clipping on a clipped video file, according to volume information in the clipped video file; and
determining, by the video clipping device, a title of the secondarily clipped video file, according to a bullet comment in the secondarily clipped video file, comprising:
transforming respective bullet comments appearing in the secondarily clipped video file into corresponding text vectors;
determining, for each text vector, a sum of distances between the text vector and other text vectors; and
taking the bullet comment corresponding to a text vector with a minimum sum of distances as the title of the secondarily clipped video file.