US 12,182,717 B2 · Feature extraction with a convolutional neural network

Abstract

Artificial intelligence is an increasingly important sector of the computer industry. One of the most important applications for artificial intelligence is object recognition and classification from digital images. Convolutional neural networks have proven to be a very effective tool for object recognition and classification from digital images. However, convolutional neural networks are extremely computationally intensive thus requiring high-performance processors, significant computation time, and significant energy consumption. To reduce the computation time and energy consumption a “cone of dependency” and “cone of influence” processing techniques are disclosed. These two techniques arrange the computations required in a manner that minimizes memory accesses such that computations may be performed in local cache memory. These techniques significantly reduce the time to perform the computations and the energy consumed by the hardware implementing a convolutional neural network.

Claims12

Dependent claims

A method of extracting features from a digital image with a multi-step convolutional processing system in a memory efficient manner, said method comprising the stages of:

dividing said digital image into a set of digital image subareas,
for each digital image subarea in said set of digital image subareas, performing the substages of:
loading a subset of digital image subarea data for a feature extraction operation,
executing said feature extraction operation on said subset of digital image data and any saved intermediate results to generate a first set of results,
executing any additional later step operations of said multi-step convolutional processing system using said first set of results of convolutional stride calculation to generate a second set of results,
saving intermediate data in said first set of results and said second set of results still needed for calculating later step operations in said multi-step convolutional processing system as saved intermediate results and discarding all other data, and
repeating said substages of loading, executing, executing, and saving until a final output from said multi-step convolutional processing system is complete for said digital image subarea;
and
combining said final outputs for each said digital image subareas into a final result for said digital image.

6 more elements

Dependencies are inferred from the claim text.

Feature extraction with a convolutional neural network

Life of this patent

Feature extraction with a convolutional neural network

Life of this patent

Abstract

Claims12