This repository contains the data analysis files and the research report for a study investigating how different levels of foveation affect perceived visual quality in Virtual Reality (VR).
Important note This was a coursework project which mainly involved using Unity to run user studies to study different levels of foveation. The core codebase is authored by the creator of the original repository.
My contribution focuses on the experimental design adaptation, data collection, data analysis, interpretation of results, and report writing.
Foveation is a VR rendering technique that reduces resolution in the user’s peripheral vision while maintaining high resolution in the central (foveal) region, leveraging properties of the human visual system to improve computational efficiency without significantly degrading perceived visual quality.
This project evaluates how different levels of foveation affect subjective visual experience across multiple VR scenes with varying visual characteristics (e.g., forests, snow landscapes, night scenes, architectural environments).
- Hardware: Varjo VR-3 headset with eye-tracking
- Platform: Unity-based VR environment
- Participants: 14 users
- Stimuli:
- 5 omnidirectional 16K images
- Mix of HDRi images and custom DSLR-captured scenes
- Conditions:
- 10 foveation levels
- Randomized presentation
- Data Collected:
- Subjective quality ratings
- Gaze tracking data
- Head rotation data
- Scene-dependence: Sensitivity to foveation strongly depends on scene content
- High-detail environments (e.g. forests, complex textures):
→ Very low tolerance to foveation - Uniform environments (e.g. snow landscapes):
→ Higher tolerance, and in some cases slight foveation was preferred - Even minimal peripheral resolution reduction was perceptible in many scenes
- Visual attention consistently concentrated near the horizon and high-texture regions
- Adapted experimental design from prior research
- Dataset construction (HDRi selection + DSLR omnidirectional capture)
- Experiment execution using Unity and Varjo VR3 headset
- Data processing & visualization
- Statistical analysis
- Interpretation of results
- Report writing
Data and Analysis Files/– Data files and analysis filesreport/– Report (PDF)
- Dynamic, gaze-contingent adaptive foveation
- Scene-aware foveation models
- Applications in gaming, training, education, and XR
- Extension to mixed reality environments
Dipayan Chowdhury
Data analysis, experimental execution, and research reporting