University of Illinois Urbana-Champaign

DeepFusion VIO: Deep-Learned Sensor Fusion for Robust Visual-Inertial Odometry

Lee, Sukwon

Permalink

https://hdl.handle.net/2142/130503

Description

Title
DeepFusion VIO: Deep-Learned Sensor Fusion for Robust Visual-Inertial Odometry
Author(s)
Lee, Sukwon
Issue Date
2025-12-10
Keyword(s)
  • Visual Inertial Odometry (VIO)
  • Deep Learning
  • Artificial Intelligence (AI)
  • Virtual Reality (VR)
  • Extended Reality (XR)
  • Augmented Reality (AR)
Date of Ingest
2025-12-10T16:18:32-06:00
Abstract
This poster is Crop Sciences Capstone Project Poster Symposia (2025) submission that presents DeepFusion VIO, a compact deep model that fuses 200 Hz IMU and dense optical flow to predict metric body-frame linear and angular velocities, which are then integrated to reconstruct 6-DoF trajectories on the EuRoC MAV Machine Hall sequences. The goal is to develop and evaluate a deep-learning-based, XR-oriented metric VIO model (DeepFusion) that is more robust than a simple ORB feature visual-odometry baseline on the EuRoC Machine Hall sequences.
Publisher
Department of Electrical and Computer Engineering, Grainger College of Engineering, University of Illinois at Urbana-Champaign
Has Part
https://colab.research.google.com/drive/1LZXC65DlAbaRXymiY6DZpzfylQbDGhsk
Series/Report Name or Number
Number Crop Sciences Student Poster Symposium, Fall 2025
Type of Resource
  • text
  • still image
Genre of Resource
conference poster
Language
eng
DOI
https://hdl.handle.net/2142/130503
Copyright and License Information
Copyright 2025 owned by Sukwon (Mike) Lee

Owning Collections

Illinois Research and Scholarship (Open Collection) PRIMARY
This is the default collection for all research and scholarship developed by faculty, staff, or students at the University of Illinois at Urbana-Champaign
Crop Sciences Capstone Project Poster Symposia