Understanding trained CNNs by indexing neuron selectivity
This work addresses the black-box nature of CNNs for researchers and practitioners in computer vision by providing automated tools to analyze neuron selectivity, though it is incremental as it builds on existing visualization and selectivity concepts.
The authors tackled the problem of understanding the internal representations of trained Convolutional Neural Networks (CNNs) by proposing a framework to index neuron selectivity for features like color and class membership, enabling the identification of specific neurons such as a red-mushroom neuron in layer Conv4 or dog-face neurons in layer Conv5 in VGG-M.
The impressive performance of Convolutional Neural Networks (CNNs) when solving different vision problems is shadowed by their black-box nature and our consequent lack of understanding of the representations they build and how these representations are organized. To help understanding these issues, we propose to describe the activity of individual neurons by their Neuron Feature visualization and quantify their inherent selectivity with two specific properties. We explore selectivity indexes for: an image feature (color); and an image label (class membership). Our contribution is a framework to seek or classify neurons by indexing on these selectivity properties. It helps to find color selective neurons, such as a red-mushroom neuron in layer Conv4 or class selective neurons such as dog-face neurons in layer Conv5 in VGG-M, and establishes a methodology to derive other selectivity properties. Indexing on neuron selectivity can statistically draw how features and classes are represented through layers in a moment when the size of trained nets is growing and automatic tools to index neurons can be helpful.