Hello
I short. I can make a serious try, since I worked with SVD/PCA in application for images, and also "played" quite a lot with simple NNs for grayscale images. But you have to understand, that working with images if somethin really simple is not enough, will require Conv and MaxPool or AveragePool layers too. At least basic, available in all ML libraries.
And - my price is much more realistic if you need serious attention, not an application of the couple of well-known patterns.
I have GPU, so for speed I can train and validate much faster.