Abstract: A computer implemented method includes obtaining data for raw image frames captured by a moving camera. The raw image frames are indexed geographically, and a graph is created from the multiple raw image frames. The graph includes image frames as vertices and edges that represent image frames having overlapping image information. The method further includes skipping frames based on the amount of overlap, determining a frame having an interesting feature, using the graph to find additional raw image frames that have the interesting feature, combining multiple raw image frames to form a unique image frame, and transmitting the unique image frame.

Abstract: A computer-implemented method to improve scale consistency and/or scale awareness in a model of self-supervised depth and ego-motion prediction neural networks processing a video stream of monocular images, wherein complementary GPS coordinates synchronized with the images are used to calculate a GPS to scale loss to enforce the scale-consistency and/or -awareness on the monocular self-supervised ego-motion and depth estimation. A relative weight assigned to the GPS to scale loss exponentially increases as training progresses. The depth and ego-motion prediction neural networks are trained using an appearance-based photometric loss between real and synthesized target images, as well as a smoothness loss on the depth predictions.

Abstract: This application generally relates to defining, displaying and interacting with tags in a 3D model. In an embodiment, a method includes generating, by a system including a processor, a three-dimensional model of an environment based on sets of aligned three-dimensional data captured from the environment, and associating tags with defined locations of the three-dimensional model, wherein the tags are respectively represented by tag icons that are spatially aligned with the defined locations of the three-dimensional model as included in different representations of the three-dimensional model rendered via an interface of a device, wherein the different representations correspond to different perspectives of the three-dimensional model, and wherein selection of the tag icons causes the tags respectively associated therewith to be rendered at the device.