US 11,866,055 B1
Tuning layers of a modular neural network
Sharan Srinivasan, Sunnyvale, CA (US); Brian Tuan, Cupertino, CA (US); John Bicket, Burlingame, CA (US); Jing Wang, Toronto (CA); Muhammad Ali Akhtar, Chicago, IL (US); Abner Ayala Acevedo, Orlando, FL (US); Bruce Kellerman, Atlanta, GA (US); and Vincent Shieh, San Francisco, CA (US)
Assigned to Samsara Inc., San Francisco, CA (US)
Filed by Samsara Inc., San Francisco, CA (US)
Filed on May 9, 2022, as Appl. No. 17/662,622.
Application 17/662,622 is a continuation of application No. 17/454,790, filed on Nov. 12, 2021, granted, now 11,352,014.
This patent is subject to a terminal disclaimer.
Int. Cl. B60W 40/09 (2012.01); G06N 3/082 (2023.01); G06V 40/20 (2022.01); G06N 3/045 (2023.01)
CPC B60W 40/09 (2013.01) [G06N 3/045 (2023.01); G06N 3/082 (2013.01); G06V 40/20 (2022.01); B60W 2400/00 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A vehicle device comprising:
a computer readable storage medium having program instructions embodied therewith; and
one or more processors configured to execute the program instructions to cause the vehicle device to:
access metadata associated with a vehicle, wherein the metadata comprises one or more characteristics of the vehicle;
compare the metadata with segmentation data, wherein the segmentation data identifies statistics associated with tuning particular models;
determine how to segment a neural network based at least in part on comparing the metadata with the segmentation data;
generate a modular neural network based at least in part on determining how to segment the neural network, the modular neural network comprising a plurality of models, wherein each of the plurality of models is independently trainable;
obtain sensor data associated with the vehicle;
execute the modular neural network based at least in part on the sensor data;
identify a particular model of the plurality of models based at least in part on an output of the modular neural network; and
independently train the particular model based at least in part on the output of the modular neural network.