I need to test the performance of CNNs with default filters against CNNs with custom filters. But I don’t know what values to choose for custom filters and I couldn’t find any source that explains my problem. I know that there a some filters like sobel and laplacian. But they don’t work on my problem.
My problem is to classify small grayscale images. These 9×9 2-dim arrays contain values between 0 and 255 or 0 and 1 (after normalization). And let’s say I want the filter to detect a corner. Let the example image look something like this:

I am working on binary classification of background and signal images using CNN. Background image comprises of white gaussian noise 2D map with underlying weak 2 point correlation. Signal image is a background image with linearly added small localized defect(s) which is well hidden under the background fluctuation. Such defects are not visible to the naked eye but have a distinct profile. Task for now is simply if network can detect images with hidden defects.