Door Detection and Distance Estimation in Habitat Simulator for 3D Environments

Doors play a pivotal role in navigation and interaction within 3D environments, whether in robotics, gaming, or architectural simulations. Accurate door detection and distance estimation are crucial for enhancing spatial awareness and enabling efficient path planning. In this research, we present a comprehensive approach to door detection and distance estimation within the Habitat simulator. Our methodology leverages the YOLOv8 deep learning model to detect doors and utilizes depth maps for precise distance estimation. We address the challenge of distinguishing between open and closed doors and demonstrate the effectiveness of our approach through a systematic evaluation. The results showcase a remarkable capability in detecting doors with high accuracy, providing valuable insights for spatial understanding and robotic navigation. With an achieved F1-score of 96.53% in our experiments, this work significantly contributes to the field of door detection and distance estimation for 3D environments, with potential applications in various domains, including robotics, virtual reality, and architectural simulations.