Method and device for economizing computing resources to be used during a process of verification of convolutional parameters using test pattern to enhance fault tolerance and fluctuation robustness in extreme situations

A method for verifying an integrity of one or more parameters of a convolution unit by inserting at least one test pattern into at least one integrated image generated from at least one input image, comprising steps of:

• (a) using a computing device, if the input image is acquired, instructing a target region estimating network to estimate a first target region to an n-th target region on the input image or its one or more resized images, wherein one or more corresponding target objects are estimated as located on each of the first target region to the n-th target region;
• (b) the computing device instructing at least one image-manipulating network to generate a first manipulated image to an n-th manipulated image, from the input image or its resized images, each of which corresponds to each of the first target region to the n-th target region and to generate the integrated image by concatenating the first manipulated image to the n-th manipulated image, wherein, at the time of the concatenation, the test pattern is inserted between at least some adjacent manipulated images among the first manipulated image to the n-th manipulated image;
• (c) the computing device instructing the convolution unit to generate at least one feature map by applying the convolution operations to the integrated image;
• and
• (d) the computing device instructing at least one verification unit to verify the integrity of the convolution unit by referring to values included in at least one area, corresponding to the test pattern, on the feature map and values included in at least one output for reference corresponding to a result of one or more operations applied to the test pattern by the convolution unit.

A computing device verifying an integrity of one or more parameters of a neural network by inserting at least one test pattern into at least one background area of at least one input image, comprising:

• at least one memory that stores instructions;
• and
• at least one processor configured to execute the instructions to perform processes of:
• (I) if information on each of a (t−2)-th image and a (t−1)-th image respectively corresponding to (t−i2)-th frame and (t−i1)-th frame of a video is acquired, instructing at least one background prediction unit to generate t-th background prediction information of a t-th image by referring to the information on said each of the (t−2)-th image and the (t−1)-th image, (II) if the t-th image is acquired, instructing at least one pattern insertion unit to insert the test pattern into the t-th image by referring to the t-th background prediction information, to thereby generate at least one input for verification, (III) instructing the neural network to generate at least one output for verification by applying one or more convolution operations to the input for verification, and (IV) instructing at least one verification unit to determine the integrity of the neural network by referring to the output for verification and at least one output for reference corresponding to the test pattern,
• wherein t, i1 and i2 are integer values.

A computing device for verifying an integrity of one or more parameters of a convolution unit by inserting at least one test pattern into at least one integrated image generated from at least one input image, comprising:

• at least one memory that stores instructions;
• and
• at least one processor configured to execute the instructions to perform processes of:
• (I) if the input image is acquired, instructing a target region estimating network to estimate a first target region to an n-th target region on the input image or its one or more resized images, wherein one or more corresponding target objects are estimated as located on each of the first target region to the n-th target region, (II) instructing at least one image-manipulating network to generate a first manipulated image to an n-th manipulated image, from the input image or its resized images, each of which corresponds to each of the first target region to the n-th target region and to generate the integrated image by concatenating the first manipulated image to the n-th manipulated image, wherein, at the time of the concatenation, the test pattern is inserted between at least some adjacent manipulated images among the first manipulated image to the n-th manipulated image, (III) instructing the convolution unit to generate at least one feature map by applying the convolution operations to the integrated image, and (IV) instructing at least one verification unit to verify the integrity of the convolution unit by referring to values included in at least one area, corresponding to the test pattern, on the feature map and values included in at least one output for reference corresponding to a result of one or more operations applied to the test pattern by the convolution unit.