A scenic mountain landscape in Rocky Mountain National Park under a vast blue sky dotted with fluffy white clouds. The sun is visible on the upper right corner, casting a bright glare. A rugged mountain slope occupies the foreground, consisting mostly of rocky terrain with sparse vegetation, leading up to a ridge line that meets the sky. In the middle distance, the mountains continue to roll with varying shades of brown and hints of green, indicating sparse grasses or low bushes. In the far distance, a series of mountain ranges extend to the horizon, suggesting a vast wilderness area. It's a clear day, and the overall feel is of open space and natural beauty typical of a high-altitude environment.
It's a picture of me! A white male wearing a bright yellow button up. I am smiling and have a pretty nice mustache. My hair is brown.

Josh Myers-Dean

Ph.D. Student

Computer Science Department

University of Colorado Boulder

Email: first.last@colorado.edu

Office: DLC 170K

Research Keywords: image segmentation, human-in-the-loop computer vision, computer vision

I am happy to chat about applying to graduate school/GRFP -- feel free to reach out!

Who am I?

I am a fourth-year Computer Science Ph.D. student at the University of Colorado Boulder, working on image segmentation and consistency in foundation models. My research focuses on automating tedious tasks in image editing applications, such as selection, and developing efficient models for use in accessibility and creative applications. I am honored to be supported by the National Science Foundation through a GRFP Fellowship and to be advised by Danna Gurari. I also frequently collaborate with the talented folks at Adobe Research.

In the past, I’ve had the privilege of interning at the Allen Institute for AI, where I worked on unsupervised learning for remote sensing time series change detection with Favyen Bastani and Aniruddha Kembhavi. I also spent time at Adobe Research, working on interactive segmentation with Brian Price, Yifei Fan, Kangning Liu, and Jason Kuen. Additionally, I had a great experience at the Pacific Northwest National Laboratory, working on natural language processing problems and VR applications related to national security and biosurveillance with Karl Pazdernik.

Before my Ph.D., I earned my Bachelor's degree (with a brief stint in studying social work) at Western Washington University, where I worked on Computational Photography and Geospatial Computer Vision with Scott Wehrwein and Structural Bioinformatics with Filip Jagodzinski. During my time there, I also worked as a web application developer, focusing on making websites more accessible and compliant with WCAG 2.0 guidelines.

Outside of my work, I embrace the Boulder stereotypes through trail and ultra running, climbing, cooking, knitting, and biking.

News

  • 2024-11: I will be joining Meta Reality Labs next year as a research intern working with Michael Iuzzolino!
  • 2024-07-01: 1 paper accepted to ECCV 2024!
  • 2024-04-17: Returning to Adobe Research this Summer!
  • 2024-04-17: The project page is live for DIG, which I presented at WACV in Hawaii. Check out the project page for the paper, code, data and supplemental material!
  • 2024-04-15: Gave a talk about some of our recent work at the CMU HCII Accessibility Lunch! Thanks for having me, Jeff and Peya!
  • 2024-04-11: Passed my Area Exam!
  • 2024-04-10: 1 paper accepted to DEF-AI-MIA.

Publications

For a complete list of my work, please check out my Google or Semantic Scholar.

A cougar with subpart decompositions.

SPIN: Hierarchical Segmentation with Subpart Granularity in Natural Images

Josh Myers-Dean, Jarek Reynolds, Brian Price, Yifei Fan, Danna Gurari

European Conference on Computer Vision (ECCV) 2024

Project Page Paper (arXiv) API (Github) Supplemental Material (46MB PDF)

Photos showcasing COVID-19 tests and their associated localizations for the result window and test iself.

Interpreting COVID Lateral Flow Tests' Results with Foundation Models

Stuti Pandey, Josh Myers-Dean, Jarek Reynolds, Danna Gurari

CVPR DEF-AI-MIA 2024

Project Page Paper (arXiv)

A schematic showing the proposed gesture-agnostic context-free interactive segmentation task. The left column shows a human, a turtle (ROI), and an optional previous segmentation. The middle column shows that different gesture types can be used, such as clicks or scribbles. The right shows a final segmentation.

Interactive Segmentation for Diverse Gesture Types Without Context

Josh Myers-Dean, Yifei Fan, Brian Price, Wilson Chan, Danna Gurari

IEEE Winter Conference on Applications of Computer Vision (WACV) 2024

Project Page Paper (3 MB PDF) Code (Github) Supplemental Material

A 3x3 grid showing overhead imagery with the country borders drawn on.

Computer Vision for International Border Legibility

Trevor Ortega, Thomas Nelson, Skyler Crane, Josh Myers-Dean, and Scott Wehrwein

IEEE Winter Conference on Applications of Computer Vision (WACV) 2023

Project Page Paper (5.1 MB PDF) Code (Github) Supplemental Material

Visualization of triplet loss on neural network features.

Generalized few-shot semantic segmentation: All you need is fine-tuning

Josh Myers-Dean, Yinan Zhao, Brian Price, Scott Cohen, Danna Gurari

arXiv preprint

Paper (8 MB PDF)

The pipeline for the DAMM software. It shows the workflow of a user giving a FASTA file which is then matched against 272 PDZ domains.

Domain Analysis and Motif Matcher (DAMM): A Program to Predict Selectivity Determinants in Monosiga brevicollis PDZ Domains Using Human PDZ Data

Haley A Wofford, Josh Myers-Dean, Brandon A Vogel, Kevin Alexander Estrada Alamo, Frederick A Longshore-Neate, Filip Jagodzinski, Jeanine F Amacher

Molecules

Paper (3 MB PDF)

A table of different computational notebooks and their benefits.

Towards Modeling Student Engagement with Interactive Computing Textbooks: An Empirical Study

David H Smith IV, Qiang Hao, Christopher D Hundhausen, Filip Jagodzinski, Josh Myers-Dean, Kira Jaeger

Proceedings of the 52nd ACM Technical Symposium on Computer Science Education

Paper (1.5 MB PDF)

An image representing the interaction of a ligand, labeled AZZ, with amino acid residues of a protein or enzyme. The ligand AZZ is in the center, with various molecular interactions illustrated by lines connecting it to surrounding residues named by their three-letter amino acid codes and position numbers (like ILE 29A, VAL 54A, etc.). There's a legend on the left side with two sections: Residue Styles and Interaction Styles. Residue Styles has icons representing side-chains and backbone & side-chain. Interaction Styles uses colored circles to denote different types of interactions: ligand acceptor (purple), non-ideal ligand acceptor (blue), ligand donor (pink), contact only (white), and stacking (green) Amino acid residues are shown with these colored circles based on their interaction with the ligand. For example, TRP 57A shows a stacking interaction (green circle) with the ligand. There are also dashed lines indicating hydrogen bonds and dotted lines indicating other non-covalent interactions. The molecular structure of AZZ is highlighted with atoms and bonds: double lines for double bonds, and a zigzag structure representing a chemical ring or chain.

PETRA: Drug Engineering via Rigidity Analysis

Sam Herr, Josh Myers-Dean, Hunter Read, Filip Jagodzinski

Molecules

Paper (1MB PDF)

An image of a flower (left) with a corresponding energy map based off semantic features and edge weights (right).

Semantic Pixel Distances for Image Editing

Josh Myers-Dean, Scott Wehrwein

NTIRE 2020 CVPR Workshop (Oral Presentation)

Project Page Paper (5MB PDF) Video (Youtube)