InstancePose: Fast 6DoF Pose Estimation for Multiple Objects From a Single RGB Image

Lee Aing, Wen-Nung Lie, Jui-Chiu Chiang, Guo-Shiang Lin; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2021, pp. 2621-2630

6DoF object pose estimation depends on positional accuracy, implementation complexity and processing speed. This study presents a method to estimate 6DoF object poses for multi-instance object detection that requires less time and is accurate. The proposed method uses a deep neural network, which outputs 4 types of feature maps: the error object mask, semantic object masks, center vector maps (CVM) and 6D coordinate maps. These feature maps are combined in post processing to detect and estimate multi-object 2D-3D correspondences in parallel for PnP RANSAC estimation. The experiments show that the method can process input RGB images containing 7 different object categories/ instances at a speed of 25 frames per second with competitive accuracy, compared with current state-of-the-art methods, which focus only on some specific conditions.