Towards Visually Explaining Variational Autoencoders
Wenqian Liu, Runze Li, Meng Zheng et al. · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · 2020
Recent advances in Convolutional Neural Network (CNN) model interpretability have led to impressive progress in visualizing and understanding model predictions. In particular, gradient-based visual attention methods have driven much recent effort in using visual attention maps as a means for visual explanations. A key problem, however, is these methods are designed for classification and categorization tasks, and their extension to explaining generative models, e.g., variational autoencoders (VAE) is not trivial. In this work, we take a step towards bridging this crucial gap, proposing the first technique to visually explain VAEs by means of gradient-based attention. We present methods to generate visual attention from the learned latent space, and also demonstrate such attention explanations serve more than just explaining VAE predictions. We show how these attention maps can be used to localize anomalies in images, demonstrating state-of-the-art performance on the MVTec-AD dataset. We also show how they can be infused into model training, helping bootstrap the VAE into learning improved latent space disentanglement, demonstrated on the Dsprites dataset.
Cited by 283 works. doi:10.1109/cvpr42600.2020.00867 Open-access PDF
Authors
References
- Deep Residual Learning for Image Recognition
- ImageNet classification with deep convolutional neural networks
- Densely Connected Convolutional Networks
- Visualizing and Understanding Convolutional Networks
- GAN(Generative Adversarial Nets)
- Generative Adversarial Nets
- Image-to-Image Translation with Conditional Adversarial Networks
- Auto-Encoding Variational Bayes
- Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization
- Learning Deep Features for Discriminative Localization