Date of Completion

Summer 5-1-2026

Thesis Advisor(s)

Joseph Johnson, Yufeng Wu

Honors Major

Computer Science

Disciplines

Computer Engineering

Abstract

This project focuses on the design and development of an AI-assisted frame selection system that automatically analyzes sequences of sports images and ranks frames based on overall photographic and contextual value. Additionally, we will focus on providing insights into the Computer Vision and Artificial Intelligence techniques used to build it.

The application utilizes computer vision and machine learning techniques to evaluate multiple dimensions of image quality and content. These include technical image quality metrics such as sharpness, motion blur, and exposure; compositional metrics such as subject placement and visual balance; and semantic understanding through detection of players, ball location, pose, and key sports actions. The project also explores higher-level visual signals such as emotional expression, peak action moments, and scene importance within game context. These signals are combined into multi-modal image quality metrics that generate an overall ranking score for each frame.

This work investigates and compares multiple computer vision approaches for frame evaluation, including object detection, pose estimation, image quality assessment models, and feature embedding techniques. Performance is evaluated using quantitative ranking metrics and qualitative comparison against human-selected highlight frames. The project also includes the development of an end-to-end prototype application that allows users to upload image sequences and receive ranked galleries with associated tags and confidence scores.

The goal of this project is to demonstrate how AI can assist creative workflows by reducing manual review time while preserving the photographer’s control over final image selection. Beyond sports photography, this work contributes to broader research in automated image curation, multi-modal visual quality assessment, and applied computer vision systems for real-world creative domains.

Accessibility Requirements

1

Download

Share

COinS