Who danced better? ranked tiktok dance video dataset and pairwise action quality assessment method
نویسندگان
چکیده
Video-based action quality assessment (AQA) is a non-trivial task due to the subtle visual differences between data produced by experts and non-experts. Current methods are extended from recognition domain where most based on temporal pattern matching. AQA has additional requirements order tempo matter for rating of an action. We present novel dataset ranked TikTok dance videos, pairwise method predicting which video same-label pair was sourced better dancer. Exhaustive pairings videos were randomly assigned 100 human annotators, ultimately producing list per label category. Our relies successful detection subject’s 2D pose inside successive query frames actions encoded String sequence. The detected returns top-matching Visual word Codebook represent current frame. Given pair, we generate value concatenated words each video. By computing edit distance score Gold Standard’s (i.e., top-ranked video(s) that category), declare with lower as winner. implemented using two schemes, i.e., without text compression. Although average precision both schemes over 12 categories low, at 0.45 compression 0.48 without, values several comparable past methods’ (median: 0.47, max: 0.66).
منابع مشابه
Balance Assessment in Children with Down Syndrome Who Practice Dance
Background. It is observed that dance delivers a valuable contribution to the development of bodily motor functions. Objectives. To evaluate the static and dynamic balance in children with Down syndrome - DS enrolled in a Dance Workshop for people with disabilities. Methods. Balance tests proposed by Oliveira (2009) were used on 14 children with DS aged between 8 and 12 (mean 9.36 ± 1.2 years...
متن کاملChina–Autos: Haven’t We Danced this Dance Before?
Just as it had in several recent similar disputes, the Panel in China– Autos found several of the challenged issues WTO-inconsistent. We believe virtually all of the deficiencies noted by the Panel could be easily addressed with minor changes to MOFCOM practices. The real significance of this dispute lies in what it tell us about the larger trade policy dance between the US and China. On the on...
متن کاملThe Kinetics Human Action Video Dataset
We describe the DeepMind Kinetics human action video dataset. The dataset contains 400 human action classes, with at least 400 video clips for each action. Each clip lasts around 10s and is taken from a different YouTube video. The actions are human focussed and cover a broad range of classes including human-object interactions such as playing instruments, as well as human-human interactions su...
متن کاملEnhanced Full-reference Quality Assessment Method for Image and Video
Evaluation and optimization of visual quality is of fundamental importance to numerous image and video processing applications including broadcasting, archiving, and restoration. As quality assessment using subjective tests is very costly and time-consuming, objective measures are required that can predict the perceptual judgment of human viewers. In this paper, a topdown full-reference quality...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: International Journal of Advances in Intelligent Informatics
سال: 2023
ISSN: ['2548-3161', '2442-6571']
DOI: https://doi.org/10.26555/ijain.v9i1.919