Amanda Duarte

Papers

Total Citations

H-Index

About

Amanda Duarte is a leading researcher in autonomous underwater robotics, with a focus on perception and navigation in challenging subsea environments. Her work bridges computer vision and deep learning to solve critical problems in underwater vehicle autonomy. Her most influential paper, "Vision-Based Obstacle Avoidance Using Deep Learning" (40 citations), introduces a novel method for Autonomous Underwater Vehicles (AUVs) to navigate safely using only a simple monocular camera. By employing a deep neural network to compute transmission maps from raw images, her approach enables real-time obstacle detection without expensive sonar equipment. Duarte has also made significant contributions to benchmarking and localization, notably through her work "Towards comparison of underwater SLAM methods: An open dataset collection" (21 citations), which provides the research community with standardized simulated datasets for evaluating simultaneous localization and mapping algorithms under varying turbidity conditions. Additionally, her research on topological descriptors for forward-looking sonar images (7 citations) advances the ability of robots to recognize previously visited locations—a key challenge for long-term autonomous missions. Through these contributions, Duarte has established herself as a key figure in making underwater robots more capable, reliable, and accessible for critical tasks like inspection, monitoring, and maintenance.