Fig. 9From: 3D human pose estimation model using location-maps for distorted and disconnected images by a wearable omnidirectional cameraHeatmaps and location-maps for input images in the same success and failure cases of Fig. 8. The maps are shown of only worst joint in the failure examples, and the right hand joint maps are shown in the success. The ground truth of location-maps are uniform distributions of x, y, and z values of the joint position. The measures of value for heatmaps and location-maps are shown below the table. Note that VNect maps are displayed on a double scale for easy comparisonBack to article page