VividVR: Real-Time Optimization of Dynamic Light Probe Placement for Enhanced Visual Fidelity in VR Environments; Using SVM and Random Forest under Changing Conditions

This paper explores optimizing light probe placement in virtual reality (VR) environments to improve visual realism and performance. We develop a methodology to evaluate light probe configurations by analyzing light intensity, color, and shadow data. Our approach includes three steps: removing probes iteratively to collect data, applying machine learning techniques to this data, and optimizing probe configurations based on the findings. We introduce Light Intensity Variance and Color Consistency as key metrics, utilize Random Forests to set optimal thresholds, and employ a Support Vector Machine classifier to automate probe configuration selection. This method reduces manual adjustments, accelerates VR content development, and enhances scene realism through strategic light probe arrangement.