Systems and methods for 3D image distification
摘要
說明書
In various embodiments, a computing device, such as a camera, sensor or scanner, can capture, generate, and/or store imagery data, such as 2D or 3D imagery data associated with an environment, setting, or for a particular purpose for which the 2D or 3D imagery is to be used.
The 2D or 3D imagery data can be used to train a predictive model, for example, via machine learning. The predictive model may be trained in a variety of machine learning techniques, such as inputting the 2D or 3D imagery into a neural network using deep learning techniques.
In some embodiments, the predictive model can be used to classify and determine driver behavior. In such an embodiment, 2D or 3D images and data of a driver captured or generated from cameras, sensors or other devices within a vehicle can be used as input into the predictive model. The model could return as output an indication or classification of one or more driver behaviors that can include, for example, “calling,” (using the right hand or the left hand), “texting” (using the right hand or left hand), “eating,” “drinking,” “adjusting the radio,” or “reaching the backseat.” A driver behavior of “normal” may also be identified, for example, if the driver has both hands on the steering wheel, one hand on the steering wheel and another on a stick-shift, etc. It is noted that, other driver behaviors, actions, or features are contemplated by the present disclosure and are not limited to the above examples.
