Centurion Patentors as of April 8, 2025

Patent №:  12270888 – Infrastructure triggering techniques to facilitate secure ultra-wideband (UWB) ranging. In one approach, a mobile device receives UWB ranging parameters over a non-UWB radio and is triggered via a second radio to perform ranging with a UWB anchor. In another, the device obtains ranging parameters and instructions for a geographic area and performs UWB ranging with a target anchor based on that data.

Patent №:  12271168 – Systems for monitoring and managing industrial settings. A system includes a backend and a sensor kit for monitoring an industrial setting. The sensor kit has an edge device and multiple sensors connected via a self-configuring network. Sensors capture data, create reporting packets using a first protocol, and send them to the edge device. The edge device forwards sensor packets to the backend over a public network. The backend processes and stores the sensor data.

Patent №:  12272127 – Detecting digital objects and generating object masks on device. Systems and methods generate object masks in digital images using a detection-masking neural network pipeline. Detection heads identify objects by class, applying multi-scale synchronized batch normalization. A masking head generates masks for detected objects, and post-processing filters out low-quality results.

Patent №:  12271983 – Generating unified embeddings from multi-modal canvas inputs for image retrieval. Systems and methods enable related image search and modification via a unified interface. A digital image and search input—potentially multi-modal (image + text)—are processed to retrieve results using a weighted query or input embedding. Retrieved results can be used to modify the original image.

Patent №:  12271804 – Wire segmentation for images using machine learning. Embodiments perform wire segmentation in images using machine learning. A first neural network generates a global probability map to identify wire-like regions. For each region, it combines the region with the global map and uses a second neural network to create a local probability map. Local maps are then aggregated for final wire detection.