Systems and methods for 3D image distification
摘要
說(shuō)明書(shū)
Distification can be performed, for example, as a preprocessing technique for a variety of applications, including, for example, interoperating 3D with 2D imagery used for predictive models. In various embodiments disclosed herein, the generation and use of ensemble systems and methods are described that provide an enhanced ensemble predictive model by combining predictions and classifications from 2D prediction and 3D prediction models. An ensemble predicative model can produce more accurate predictions than the 2D or 3D image models alone. For example, in a test set of over 70,000 sample images depicting driver behavior, an ensemble prediction model correctly classified 96.9% of the images, whereas a stand-alone 3D CNN model and a stand-alone 2D CNN model were only able to correctly classify the same set of sample images with 93.9% and 86.1% accuracy, respectively.
As described herein, an ensemble model may use pairs of 2D and 3D images, where the pair of images are taken of the same object, scene or otherwise relate to the same frame. For example, 2D and 3D camera(s) or other computing device, for example, the computing devices disclosed for
該功能需要專(zhuān)業(yè)版企業(yè)版VIP權(quán)限，您可以：
掃描加專(zhuān)利匯官方微信交流群
