ShapeShop: Towards Understanding Deep Learning Representations via Interactive Experimentation

Fred Hohman, Nathan Hodas, Duen Horng (Polo) Chau

The ShapeShop system user interface is divided into two main sections. The Model Builder (top) contains the training data, model, and hyperparameter selection options where a user follows enumerated steps, concluding with the system building and training an N-image classifier, where each training image selected corresponds to one class. In the Experiment Results section (bottom), each time the "Train and Visualize" button is clicked a new set of results appears including the class activation maximization of each class.

Abstract

Deep learning is the driving force behind many recent technologies; however, deep neural networks are often viewed as “black-boxes” due to their internal complexity that is hard to understand. Little research focuses on helping people explore and understand the relationship between a user’s data and the learned representations in deep learning models. We present our ongoing work, ShapeShop, an interactive system for visualizing and understanding what semantics a neural network model has learned. Built using standard web technologies, ShapeShop allows users to experiment with and compare deep learning models to help explore the robustness of image classifiers.

Citation

ShapeShop: Towards Understanding Deep Learning Representations via Interactive Experimentation
Fred Hohman, Nathan Hodas, Duen Horng (Polo) Chau
Extended Abstracts on ACM Human Factors in Computing Systems (CHI). Denver, CO, USA, 2017.
Project PDF Video Poster Code BibTeX

BibTeX

@inproceedings{hohman2017shapeshop,
  title={ShapeShop: Towards Understanding Deep Learning Representations via Interactive Experimentation},
  author={Hohman, Fred and Hodas, Nathan and Chau, Duen Horng},
  booktitle={Proceedings of the 2017 CHI Conference Extended Abstracts on Human Factors in Computing Systems},
  year={2017},
  publisher={ACM}
}