Vitaly Bezuglyj

Moscow Institute of Physics and Technology

Papers

Total Citations

H-Index

About

Vitaly Bezuglyj is advancing the frontier of autonomous vehicle perception through innovative work in 3D LiDAR segmentation and multimodal place recognition. His research centers on developing robust deep learning architectures that enable self-driving systems to understand complex environments. In his highly cited work, "DAPS3D: Domain Adaptive Projective Segmentation of 3D LiDAR Point Clouds" (2023, 13 citations), Bezuglyj tackles a critical bottleneck: the domain gap between synthetic training data and real-world LiDAR scans. He introduced a domain-adaptive projective segmentation framework that bridges this divide, allowing neural networks trained on labeled simulated data to generalize effectively to diverse real-world conditions—a breakthrough for scalable autonomous driving. More recently, with "MSSPlace: Multi-Sensor Place Recognition With Visual and Text Semantics" (2025), he explores how fusing camera imagery, LiDAR point clouds, and even textual semantic cues can dramatically improve a vehicle’s ability to recognize previously visited locations. This multimodal approach promises more reliable navigation in GPS-denied or visually ambiguous environments. Bezuglyj’s work directly addresses the practical challenges of deploying autonomous systems at scale, making him a rising voice in the intersection of computer vision, robotics, and 3D scene understanding.