A vision library for performing sliced inference on large images/small objects