Interpretable Convolutional Neural Network

This paper by Quanshi Zhang, Ying Nian Wu, and Song-Chun Zhu from University of California, Los Angeles proposes a method to modify traditional convolutional neural networks (CNNs) into interpretable CNNs, in order to clarify knowledge representations in high conv-layers of CNNs. Problem: without any additional human supervision, can we modify a CNN to obtain interpretable knowledge representations in its conv-layers? Bau et al. [1] defined six kinds of semantics in CNNs, i.e. objects, parts, scenes, textures, materials, and colors. In fact, we can roughly consider the first two semantics as object-part patterns with specific shapes, and summarize the last four semantics as texture patterns without clear contours. Filters in low conv-layers usually describe simple textures, whereas filters in high conv-layers are more likely to represent object parts.