Trade-off between computing consumption brought by K and performance gain.

محفوظ في:

التفاصيل البيبلوغرافية
المؤلف الرئيسي:	Lingwen Meng (8968106) (author)
مؤلفون آخرون:	Guobang Ban (20678392) (author), Guanghui Xi (8083121) (author), Siqi Guo (355869) (author)
منشور في:	2025
الموضوعات:	Biochemistry Microbiology Physiology Biotechnology Science Policy Biological Sciences not elsewhere classified Chemical Sciences not elsewhere classified three key components refined temporal interval experimental results show attracted increasing attention advanced sstal methods achieves significant improvement ablation experiments show study uses actionformer manually annotated data tal pseudo label method called pseudo localization accuracy compared pseudo label refining label refining pseudo label unlabeled data study proposes consuming compared propose pseudo label self label rate supervised method method proposed xlink "> video annotation student framework multiple inference large amount labels obtained image annotation different qualities combines labeled boundary accuracy activitynet v1 60 %. 3 datasets
الوسوم:	إضافة وسم لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!

Trade-off between computing consumption brought by <i>K</i> and performance gain.