Anna Vorontsova

About

Anna Vorontsova is a computer vision and robotics researcher whose work sits at the intersection of 3D scene understanding, autonomous navigation, and depth perception. She has made significant contributions to 3D object detection, visual SLAM, and depth estimation — areas critical to advancing robotic systems and augmented reality applications. Her most influential work, "FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection" (2022), has garnered 115 citations and introduced a streamlined, anchor-free framework that pushed the boundaries of efficiency and accuracy in 3D point cloud detection. Building on this foundation, her recent "UniDet3D" (2025) tackles a key limitation in the field — the insufficiency of individual indoor datasets — by training across multiple datasets simultaneously to produce more generalizable detectors. Vorontsova has also advanced the robotics community's understanding of visual SLAM robustness (58 citations), producing rigorous feasibility analyses for real-world indoor navigation scenarios. Her dataset contribution, DISCOMAN, further demonstrates a commitment to providing the research community with richly annotated resources for odometry and semantic mapping. Across her portfolio, Vorontsova consistently bridges theoretical innovation with practical, deployable solutions for intelligent robotic systems.

Research Focus

Key Achievements

Top Papers

1. 1
   FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection

   115 citations · 2022
2. 2
   Measuring robustness of Visual SLAM

   58 citations · 2019
3. 3
   UniDet3D: Multi-dataset Indoor 3D Object Detection

   12 citations · 2025
4. 4
   Single-Stage 3D Geometry-Preserving Depth Estimation Model Training on Dataset Mixtures with Uncalibrated Stereo Data

   5 citations · 2022
5. 5
   DISCOMAN: Dataset of Indoor SCenes for Odometry, Mapping And Navigation

   3 citations · 2019

Key Collaborators

Contact & Links