Official websites use .gov

A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS

A lock ( 🔒 ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Need Help

Connecting Disciplines: Applying Computer Vision to Physical Activity Measurement Methods

Award Number: F31LM015004
ORGANIZATION: NATIONAL LIBRARY OF MEDICINE
OPDIV: NIH
AWARD CLASS: DISCRETIONARY
AWARD ACTIVITY TYPE: FELLOWSHIP/SCHOLARSHIP/STUDENT LOANS
PERIOD OF PERFORMANCE START DATE: 02/08/2027
PERIOD OF PERFORMANCE END DATE: 02/07/2028

Group Awards By:

View Award Description

Connecting Disciplines: Applying Computer Vision to Physical Activity Measurement Methods - PROJECT SUMMARY/ABSTRACT The high prevalence of sedentary behavior in developed nations and increasing concern of its impact on public health has resulted in a growing interest in measuring sedentary behavior. Current methods of measuring sedentary behavior involve assessing human posture through direct observation, the gold-standard methodology used by physical activity (PA) measurement researchers. The current practice of direct observation is labor-intensive and inherently subjective: necessitating high amounts of time, money, and expertise. Computer vision, the process of extracting information from images using machine learning, offers a solution to reduce the time-consuming, expensive, and subjective nature of direct observation through automation, ultimately leading to more accurate and consistent measures of sedentary behavior. Estimating physical behaviors with computer vision has traditionally been done by computer scientists in highly controlled settings and rarely been tested in real-world public health applications. Consequently, the terms that computer vision researchers and PA researchers use to describe behavior are different, thus requiring explicit collaboration and mutual understanding of the respective fields to create a useful tool. In addition to benefiting the field of PA research, rigorous testing in an applied domain will benefit the field of computer vision developmentally, documenting the limitations and working state of current models. Currently, few PA researchers have applied computer vision to direct observation data due to the gap in knowledge and technical skills. Therefore, it is imperative to train interdisciplinary researchers to create practical tools for PA researchers and further the development of computer vision algorithms. Through the training aims outlined in this proposal, the applicant will refine his competency in PA measurement, computer vision, and statistics. More specifically, the applicant will develop the necessary skills to implement computer vision using Python, create GitHub repositories for transparent collaboration, and deepen his understanding of PA measurement challenges. Through achieving the training aims, the applicant will accomplish the scientific aims of this project: 1) to evaluate the performance of existing computer vision models and 2) to develop a computer vision model to improve upon existing ones. The methods used to achieve these aims involve four stages: 1) manual annotation of criterion data, 2) training and evaluating existing computer vision models, 3) detailed exploration of the limitations of the applied models, and 4) developing and evaluating a novel ensemble method. When the scientific and training aims have been attained, we will have developed an automated posture estimation method and trained an interdisciplinary and collaborative researcher prepared to develop and apply technical innovations to overcome PA measurement challenges. Both the research product and the applicant’s skillset will contribute toward the advancement of physical behavior measurement.


Issue Date FY	Funding FY	Legal Entity Name	Legal Entity Address	Legal Entity City	Legal Entity State	Legal Entity Zip Code	Legal Entity COUNTY	Legal Entity COUNTRY	Assistance Listing	Award Code	Budget Year	Action Date	Action Type	Action Amount

Issue Date FY: 2026 ( Subtotal = $40,941 )
2026	2026	UNIVERSITY OF TENNESSEE	201 ANDY HOLT TOWER	KNOXVILLE	TN	37996	KNOX	USA	Medical Library Assistance	000	1	8/10/2026	NEW	$40,941
														Subtotal = $40,941

Grand Total All Awards = $40,941

Top

All Categories

About

Search

Reports

Data Submission

Award Information

Connecting Disciplines: Applying Computer Vision to Physical Activity Measurement Methods

Award Number: F31LM015004

ORGANIZATION: NATIONAL LIBRARY OF MEDICINE

OPDIV: NIH

AWARD CLASS: DISCRETIONARY

AWARD ACTIVITY TYPE: FELLOWSHIP/SCHOLARSHIP/STUDENT LOANS

PERIOD OF PERFORMANCE START DATE: 02/08/2027

PERIOD OF PERFORMANCE END DATE: 02/07/2028

Federal Websites

Department of Health & Human Services

HHS Operating Divisions

HHS Staff Divisions

Download A Document Viewer