US 12,321,384 B2
Method and device for video moment retrieval based on selective debiasing
Changdong Yoo, Daejeon (KR); and Sunjae Yoon, Daejeon (KR)
Assigned to KOREA ADVANCED INSTITUTE OF SCIENCE AND TECHNOLOGY, Daejeon (KR)
Filed by KOREA ADVANCED INSTITUTE OF SCIENCE AND TECHNOLOGY, Daejeon (KR)
Filed on Nov. 2, 2023, as Appl. No. 18/386,311.
Claims priority of application No. 10-2022-0154621 (KR), filed on Nov. 17, 2022.
Prior Publication US 2024/0176819 A1, May 30, 2024
Int. Cl. G06F 7/00 (2006.01); G06F 16/735 (2019.01); G06F 16/783 (2019.01); G06V 20/40 (2022.01)
CPC G06F 16/735 (2019.01) [G06F 16/7837 (2019.01); G06V 20/41 (2022.01); G06V 20/46 (2022.01)] 16 Claims
OG exemplary drawing
 
1. A method for video moment retrieval performed by a computing device, comprising:
obtaining a pair of video and query to perform the video moment retrieval;
determining based on meaning of the query whether a retrieval bias regarding the query has a positive effect on the video moment retrieval to generate a determination result; and
selectively removing the retrieval bias from a result of the video moment retrieval according to the determination result to generate a final retrieval result,
wherein the determining whether the retrieval bias regarding the query has a positive effect on the video moment retrieval includes determining whether the retrieval bias has a positive effect on the video moment retrieval based on one or more decision rules including a co-occurrence table and a learnable confounder,
wherein the co-occurrence table is a table constructed by counting co-occurrence of predicates for each object word in the training query so that each row represents the frequency of co-occurrence of the predicates for each object word, and
wherein the determining whether the retrieval bias regarding the query has a positive effect on the video moment retrieval includes identifying key predicates associated with each of the object words in the query in the co-occurrence table and determining that the retrieval bias has a positive effect on the video moment retrieval when the key predicates are included in the query.