System, method, and apparatus for a neural network model for a vehicle
AssigneesVolkswagen Aktiengesellschaft, Audi AG, Porsche AG
InventorsPratik Prabhanjan Brahma, Nasim Souly, Nikhil George
Life of this patent
Estimating the term…
Abstract
A computing system comprises a data storage and at least one processor communicatively coupled to the data storage. The at least one processor is configured to execute program instructions to cause the system to perform the following steps. A deep neural network (“DNN”) model is trained using training data. Next, additional scenes are determined based on the DNN model and the training data. The determined scenes are generated, and then used to augment the training dataset. The DNN model is then retrained using the augmented training dataset and stored in a data storage for deployment.
Claims10
A computing system, comprising:
- a data storage;
- and
- at least one processor communicatively coupled to the data storage, wherein the at least one processor is configured to execute program instructions to cause the system to perform the steps comprising:
- training a deep neural network (“DNN”) model with initial training data corresponding to a set of driving scenarios;
- determining scenes to generate for augmenting the initial training data based on the DNN model and the initial training data, wherein said scenes each comprise an image or a series of images and determining said scenes comprises:
- grouping the initial training data to statistical bins of driving scenarios each bin representing a respective probability of occurrence for those driving scenarios;
- determining underrepresented statistical bins, wherein underrepresented statistical bins comprise driving scenarios that are statistically rare to occur;
- and
- associating scenes to the underrepresented statistical bins;
- generating the determined scenes;
- augmenting the initial training data with the generated scenes;
- retraining the DNN model with the augmented training data;
- and
- storing the retrained DNN model in the data storage.
A computer-implemented method for generating a deep neural network (“DNN”) model, comprising:
- training a deep neural network (“DNN”) model with initial training data corresponding to a set of driving scenarios;
- determining scenes to generate for augmenting the initial training data based on the DNN model and the initial training data, wherein said scenes each comprise an image or a series of images and determining said scenes comprises:
- grouping the initial training data to statistical bins of driving scenarios each bin representing a respective probability of occurrence for those driving scenarios;
- determining underrepresented statistical bins, wherein underrepresented statistical bins comprise driving scenarios that are statistically rare to occur;
- and
- associating scenes to the underrepresented statistical bins;
- generating the determined scenes;
- augmenting the initial training data with the generated scenes;
- retraining the DNN model with the augmented training data;
- and
- storing the retrained DNN model in a data storage.
A non-transitory computer readable medium encoded with instructions that when executed by at least one processor causes the processor to carry out the following operations:
- training a deep neural network (“DNN”) model with initial training data corresponding to a set of driving scenarios;
- determining scenes to generate for augmenting the initial training data based on the DNN model and the initial training data, wherein said scenes each comprise an image or a series of images and determining said scenes comprises:
- grouping the initial training data to statistical bins of driving scenarios each bin representing a respective probability of occurrence for those driving scenarios;
- determining underrepresented statistical bins, wherein underrepresented statistical bins comprise driving scenarios that are statistically rare to occur;
- and
- associating scenes to the underrepresented statistical bins;
- generating the determined scenes;
- augmenting the initial training data with the generated scenes;
- retraining the DNN model with the augmented training data;
- storing the retrained DNN model in a data storage;
- and
- deploying the stored DNN to one or more vehicles.
Dependencies are inferred from the claim text.