Robotic agents need to know the position and orientation of objects in their environment. NVIDIA's Deep Object Pose Estimation (DOPE) uses synthetic data to train a model for accurate pose estimation. It handles object occlusion and bridges the reality gap by using domain randomized and photorealistic synthetic data. DOPE is a one-shot DNN that provides accurate object pose estimation for robotic manipulation.

7m read timeFrom developer.nvidia.com
Post cover image
Table of contents
Deep Object Pose EstimationAdvantages of DOPEReality gap challengeArchitectureDatasetData generationObject symmetryTraining DOPEInference and evaluationUsing ‌Isaac ROS pose estimation

Sort: