Safety-Driven Path Selection Using Reinforcement Learning in Autonomous Driving
A path-selection method that evaluates crowd-sourced road reports, updates reporter trust, and balances route risk against travel time.
Published · RSAE 2025
Paper
GBC: Gaussian-Based Colorization and Super-Resolution for 3D Reconstruction
GBC restores low-resolution monochrome video with temporally consistent colorization and super-resolution before reconstructing the scene with 3D Gaussian splatting.
Published · ACM SIGGRAPH VRCAI 2024
BMS³: Bayesian Modeling Based SwinUNet Segmentation on Self-Distillation Architecture
BMS³ combines Bayesian feature decomposition, SwinUNet, and optional self-distillation to improve cross-domain medical image segmentation.
Published · ICONIP 2025
AMECxSV: Adaptive Metadata-Driven Calibration for X-Lingual Speaker Verification
AMECxSV calibrates scores from fixed speaker encoders with language and duration context, improving full-coverage discrimination and calibration across X-lingual trials.
Preprint · arXiv 2026
SPARC: Anchored Residual Calibration for Speaker Verification
SPARC adds a compact residual to a frozen context-affine anchor, calibrating speaker-verification scores with duration, VAD, and language context.
Research project
