Using Sub-Networks Created from Neural Networks for Processing Color Images

Abstract

A system comprising an electronic device that includes a processor is described. During operation, the processor acquires a full version of a neural network, the neural network including internal elements for processing instances of input image data having a set of color channels. The processor then generates, from the neural network, a set of sub-networks, each sub-network being a separate copy of the neural network with the internal elements for processing at least one of the color channels in instances of input image data removed, so that each sub-network is configured for processing a different set of one or more color channels in instances of input image data. The processor next provides the sub-networks for processing instances of input image data—and may itself use the sub-networks for processing instances of input image data.

What is claimed is: 
     
         1 . A system, comprising:
 a processor configured to:
 generate, from a neural network for processing a set of color channels in instances of input image data, a sub-network for processing a reduced set of color channels that includes fewer color channels than the set of color channels. 
   
     
     
         2 . The system of  claim 1 , wherein the processor is configured to:
 generate, from the neural network, an additional sub-network for processing an additional set of color channels that includes:
 fewer color channels than the set of color channels; and 
 different color channels than the reduced set of color channels. 
   
     
     
         3 . The system of  claim 2 , wherein the processor is configured to:
 combine the sub-network and the additional sub-network to create an ensemble network.   
     
     
         4 . The system of  claim 2 , wherein the processor is configured to:
 select the sub-network and/or the additional sub-network for processing instances of input image data based at least in part on one or more color channels to be processed.   
     
     
         5 . The system of  claim 2 , wherein the processor is configured to:
 provide representations of the sub-network and the additional sub-network to an additional processor, the representations of the sub-network and the additional sub-network configured to be combined by the additional processor into an ensemble network to be used for processing instances of input image data.   
     
     
         6 . The system of  claim 1 , wherein the processor is configured to:
 provide a representation of the sub-network to an additional processor, the representation of the sub-network configured to be used by the additional processor for processing instances of input image data.   
     
     
         7 . The system of  claim 1 , wherein generating the sub-network includes:
 creating a copy of the neural network from which internal elements for processing at least one color channel are removed.   
     
     
         8 . The system of  claim 1 , wherein the processor is configured to:
 selecting at least one color channel to be included in the reduced set of color channels.   
     
     
         9 . The system of  claim 1 , wherein the processor is configured to:
 use the sub-network for processing instances of input image data.   
     
     
         10 . The system of  claim 1 , wherein the set of color channels includes a specified set of color channels used for digital still images and/or frames of digital video. 
     
     
         11 . A method, comprising:
 generating, from a neural network for processing a set of color channels in instances of input image data, a sub-network for processing a reduced set of color channels that includes fewer color channels than the set of color channels.   
     
     
         12 . The method of  claim 11 , further comprising:
 generating, from the neural network, an additional sub-network for processing an additional set of color channels that includes:
 fewer color channels than the set of color channels; and 
 different color channels than the reduced set of color channels. 
   
     
     
         13 . The method of  claim 12 , further comprising:
 combining the sub-network and the additional sub-network to create an ensemble network.   
     
     
         14 . The method of  claim 12 , further comprising:
 selecting the sub-network and/or the additional sub-network for processing instances of input image data based at least in part on one or more color channels to be processed.   
     
     
         15 . The method of  claim 12 , further comprising:
 providing representations of the sub-network and the additional sub-network to a processor, the representations of the sub-network and the additional sub-network configured to be combined by the processor into an ensemble network to be used for processing instances of input image data.   
     
     
         16 . The method of  claim 11 , further comprising:
 providing a representation of the sub-network to a processor, the representation of the sub-network configured to be used by the processor for processing instances of input image data.   
     
     
         17 . The method of  claim 11 , wherein generating the sub-network includes:
 creating a copy of the neural network from which internal elements for processing at least one color channel are removed.   
     
     
         18 . The method of  claim 11 , further comprising:
 selecting at least one color channel to be included in the reduced set of color channels.   
     
     
         19 . The method of  claim 11 , further comprising:
 using the sub-network for processing instances of input image data.   
     
     
         20 . The method of  claim 11 , wherein the set of color channels includes a specified set of color channels used for digital still images and/or frames of digital video.