25 years of CNNs: Can we compare to human abstraction capabilities?

摘要

We try to determine the progress made by convolutional neural networks overthe past 25 years in classifying images into abstractc lasses. For this purposewe compare the performance of LeNet to that of GoogLeNet at classifyingrandomly generated images which are differentiated by an abstract property(e.g., one class contains two objects of the same size, the other class twoobjects of different sizes). Our results show that there is still work to do inorder to solve vision problems humans are able to solve without muchdifficulty.