Andra Petrovai

Papers

Total Citations

H-Index

About

Andra Petrovai is a leading researcher in computer vision for automated driving, specializing in real-time panoptic segmentation—a critical technology that fuses semantic and instance understanding of driving scenes. Her major contributions center on developing fast, efficient neural network architectures that can deliver high-quality panoptic segmentation without the computational overhead of traditional two-stage detectors. In her highly cited 2020 work, "Real-Time Panoptic Segmentation with Prototype Masks for Automated Driving" (16 citations), she introduced a fully convolutional network that treats panoptic segmentation as a dense classification problem, generating prototype masks for both "stuff" (e.g., road, sky) and "things" (e.g., cars, pedestrians) classes. This approach achieved state-of-the-art speed and accuracy, making it suitable for real-time autonomous driving applications. Building on this, her 2022 paper "Fast Panoptic Segmentation with Soft Attention Embeddings" (12 citations) further refined the method by incorporating soft attention mechanisms to improve mask quality and efficiency. Petrovai’s work has been instrumental in bridging the gap between research-grade panoptic segmentation and practical, deployment-ready systems, earning recognition for its impact on safe, real-time environment perception in autonomous vehicles.