Mono-Hydra++: Real-Time Monocular Scene Graph Construction with Multi-Task Learning for 3D Indoor Mapping
Mono-Hydra++ is a real-time pipeline for indoor mapping using only a camera and IMU, bypassing the need for expensive RGB-D or LiDAR sensors. It combines a powerful multi-task model (M2H-MX) with visual-inertial odometry to create metric 3D maps and hierarchical scene graphs. This enables agile robots to understand their environment spatially without relying on external depth sensors.