License plate recognition joins Sighthound’s vehicle make, model and color recognition in a new Vehicle Product Suite that is available to companies in fields such as law enforcement, traffic control, toll monitoring, and motor vehicle transportation. The technology draws upon Sighthound’s experience in building fast, deep Convolutional Neural Networks (CNNs) to perform complex tasks accurately with much more speed and much less data than in the past. Benchmarks are available in a white paper here. Demonstrations of the software running on an nVidia TX1 edge device and in the Sighthound Cloud are available at The deeply learned system is significantly more accurate than any open source or commercial alternatives to which Sighthound had access. The software is designed to return the state for US license plates and country for certain EU countries. The license plate detector can find license plates in a much broader range of real world deployments than traditional single lane focus cameras, and the recognition models are not only robust to license plates from many different parts of the world but also to natural variations in conditions. In addition, it also provides confidence scores for each of the above to help with custom development. The software scans images or videos and uses Sighthound’s proprietary deep computer vision algorithms to detect and read license plates as well as recognize the country/state of origin. today announced the release of a deeply learned license plate detection and recognition API in its Cloud service and as part of its Sighthound Sentry software. PALO ALTO, Calif., Ma(GLOBE NEWSWIRE) - Sighthound, Inc.
0 Comments
Leave a Reply. |