Please use this identifier to cite or link to this item:
http://arks.princeton.edu/ark:/88435/dsp01jh343w653
Title: | Transparent and Objective AI Scoring System for Competitive Diving |
Authors: | Okamoto, Lauren |
Advisors: | Rusinkiewicz, Szymon |
Department: | Computer Science |
Class Year: | 2024 |
Abstract: | Judging in diving should be as objective and fair as possible. However, the current scoring system is inherently subjective and lacks any explanation from judges regarding how they determine their scores. Divers should be awarded for all of their efforts (without the influence of bias) as well as be able to access a breakdown of how their score was determined. This work proposes and implements a system for objectively and transparently scoring platform dives. The system first uses object detection and pose estimation to abstract the necessary visual elements of the dive (namely the platform, splash, and pose of the diver), and then uses these abstracted elements to perform dive recognition and phase segmentation. Using this information, the system performs dive error analysis such that it calculates and locates these errors: feet apart, height off board, distance from board, somersault tightness, knee straightness, twist straightness, verticalness (under/over-rotation), body straightness during entry, and splash size. Each error is then scored and aggregated to come to an overall score, and a score report is generated that summarizes each error, its description, score, and visual evidence. As verified by a group of domain experts, this report may be used to assist judges in scoring, help train judges, and provide feedback to divers. The system outperforms current state-of-the-art models in the tasks of dive recognition and phase segmentation, and experts agree with the system's outputted scores at least 90% of the time for all components. This work serves as a first step to making judging in diving more fair and objective so that no diver receives a score they don't deserve. |
URI: | http://arks.princeton.edu/ark:/88435/dsp01jh343w653 |
Type of Material: | Princeton University Senior Theses |
Language: | en |
Appears in Collections: | Computer Science, 1987-2024 |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
OKAMOTO-LAUREN-THESIS.pdf | 2.35 MB | Adobe PDF | Request a copy |
Items in Dataspace are protected by copyright, with all rights reserved, unless otherwise indicated.