Multi-domain cascade convolutional neural network

At least one non-transitory machine accessible storage medium having instructions stored thereon, wherein the instructions, when executed on a machine, cause the machine to:

• receive compressed data from a first device of a plurality of devices, wherein the compressed data is associated with visual data captured by one or more sensors;
• perform a current stage of processing on the compressed data using a current convolutional neural network (CNN), wherein the current stage of processing corresponds to one of a plurality of processing stages associated with the visual data, and wherein the current CNN corresponds to one of a plurality of convolutional neural networks (CNNs) associated with the plurality of processing stages;
• obtain an output associated with the current stage of processing, wherein the output is obtained from the current CNN;
• determine, based on the output associated with the current stage of processing, whether processing associated with the visual data is complete;
• upon a determination that the processing associated with the visual data is complete, output a result associated with the visual data;
• and
• upon a determination that the processing associated with the visual data is not complete, transmit the compressed data to a second device of the plurality of devices, wherein the second device is to perform a subsequent processing stage of the plurality of processing stages.

A method, comprising:

• receiving compressed data from a first device of a plurality of devices, wherein the compressed data is associated with visual data captured by one or more sensors;
• performing a current stage of processing on the compressed data using a current convolutional neural network (CNN), wherein the current stage of processing corresponds to one of a plurality of processing stages associated with the visual data, and wherein the current CNN corresponds to one of a plurality of convolutional neural networks (CNNs) associated with the plurality of processing stages;
• obtaining an output associated with the current stage of processing, wherein the output is obtained from the current CNN;
• determining, based on the output associated with the current stage of processing, whether processing associated with the visual data is complete;
• upon a determination that the processing associated with the visual data is complete, outputting a result associated with the visual data;
• and
• upon a determination that the processing associated with the visual data is not complete, transmitting the compressed data to a second device of the plurality of devices, wherein the second device is to perform a subsequent processing stage of the plurality of processing stages.

A system, comprising:

• one or more sensors to capture visual data;
• and
• a plurality of processing devices to:
• obtain compressed data corresponding to the visual data captured by the one or more sensors;
• extract a first set of compressed features from the compressed data;
• process the first set of compressed features using a first convolutional neural network (CNN), wherein the first CNN is trained to process one or more first types of compressed features;
• extract a second set of compressed features from the compressed data;
• process the second set of compressed features using a second convolutional neural network (CNN), wherein the second CNN is trained to process one or more second types of compressed features different from the one or more first types of compressed features;
• extract the visual data from the compressed data, wherein the visual data is extracted by decompressing the compressed data;
• and
• process the visual data using a third convolutional neural network (CNN), wherein the third CNN is trained to process raw visual data.