We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computer Vision and Pattern Recognition

Title: Visual Explanations for Convolutional Neural Networks via Latent Traversal of Generative Adversarial Networks

Abstract: Lack of explainability in artificial intelligence, specifically deep neural networks, remains a bottleneck for implementing models in practice. Popular techniques such as Gradient-weighted Class Activation Mapping (Grad-CAM) provide a coarse map of salient features in an image, which rarely tells the whole story of what a convolutional neural network (CNN) learned. Using COVID-19 chest X-rays, we present a method for interpreting what a CNN has learned by utilizing Generative Adversarial Networks (GANs). Our GAN framework disentangles lung structure from COVID-19 features. Using this GAN, we can visualize the transition of a pair of COVID negative lungs in a chest radiograph to a COVID positive pair by interpolating in the latent space of the GAN, which provides fine-grained visualization of how the CNN responds to varying features within the lungs.
Comments: 2 pages, 2 figures, to appear as extended abstract at AAAI-22
Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Image and Video Processing (eess.IV)
ACM classes: I.5.4; I.5.1; I.4.9; I.2.10
Cite as: arXiv:2111.00116 [cs.CV]
  (or arXiv:2111.00116v2 [cs.CV] for this version)

Submission history

From: Amil Dravid [view email]
[v1] Fri, 29 Oct 2021 23:26:09 GMT (5365kb,D)
[v2] Tue, 2 Nov 2021 00:42:41 GMT (5365kb,D)

Link back to: arXiv, form interface, contact.