Understanding Intra-Class Knowledge Inside CNN
This work provides insights into CNN interpretability for researchers, enabling style-based applications like image retrieval and object completion, but it is incremental as it builds on existing visualization models.
The paper tackles the problem of understanding how Convolutional Neural Networks (CNNs) represent object classes internally by visualizing intra-class knowledge in fully-connected layers, using a non-parametric patch prior to invert this knowledge into interpretable images, revealing hierarchical and ensemble organization of object styles.
Convolutional Neural Network (CNN) has been successful in image recognition tasks, and recent works shed lights on how CNN separates different classes with the learned inter-class knowledge through visualization. In this work, we instead visualize the intra-class knowledge inside CNN to better understand how an object class is represented in the fully-connected layers. To invert the intra-class knowledge into more interpretable images, we propose a non-parametric patch prior upon previous CNN visualization models. With it, we show how different "styles" of templates for an object class are organized by CNN in terms of location and content, and represented in a hierarchical and ensemble way. Moreover, such intra-class knowledge can be used in many interesting applications, e.g. style-based image retrieval and style-based object completion.