Download

Download Full Text (355 KB)

Description

In many sports, videos are being used to assist in judging. For example, they are being used to review quick actions or confirm the actions being performed. For diving specifically, the videos of a dive can be at most 3 seconds long, and divers can perform a range of somersaults and twists within that time frame, all of which affect the score, classification, and dive number. Using these videos, a Multitask Learning Action Quality Assessment (ML-AQA) program, based on machine learning, from a vast dataset, is able to produce an Action Quality Score out of 100, Factorized Action Recognition (classifications), and AQA-oriented generated captions. In this research, we developed a Python program to process a short diving video and generate the AQA score, classification results, dive number, and corresponding caption. This format makes it easier to understand and eliminates the need for users to convert the video into a specific format. This also outputs the information a person would want to know in a more readable format. The MTL-AQA analysis, as shown in other papers, demonstrated better performance compared to Single-task Learning Action Quality Assessment (STL-AQA) on the same dataset. This program can be used by divers in training to obtain scores and classifications, allowing them to improve their diving skills. Additionally, this program could be trained to work for another sport with discernible actions, such as figure skating or gymnastics.

Publisher Location

Las Vegas (Nev.)

Publication Date

Fall 11-21-2025

Publisher

University of Nevada, Las Vegas

Language

English

Keywords

Computer Science; Machine Learning; Sport Analysis; Image/Video Processing; Performance Judging

Disciplines

Electrical and Computer Engineering | Engineering

File Format

PDF

File Size

296 KB

Permissions

Google Drive\Institutional Repository\OUR_OfficeOfUGResearch\Symposia\2025 Fall Symposium

Comments

Mentor: Brendan Morris

Rights

IN COPYRIGHT. For more information about this rights statement, please visit http://rightsstatements.org/vocab/InC/1.0/

Diving Video Analysis Using Multitask Learning Action Quality Assessment (MTL-AQA)


Share

COinS