Figure 6
From: Relative pose estimation from panoramic images using a hybrid neural network architecture
Overview of the proposed neural network (left, orange) embedded in the homing architecture (left, black) and computation results (right). The example images on the right are captured from an actual forward pass through the network and homing algorithm. The input images are selected from two different illumination variants of the setting Uni Office. Size annotations on the left side are in reference to panoramic color images (\(2 \times \text {height} \times \text {width} \times \text {color channels}\)) used throughout this work. For more information on the input images, see “Dataset”.
