XRAI: Better Attributions Through Regions

@article{Kapishnikov2019XRAIBA,
  title={XRAI: Better Attributions Through Regions},
  author={A. Kapishnikov and Tolga Bolukbasi and Fernanda Vi'egas and Michael Terry},
  journal={2019 IEEE/CVF International Conference on Computer Vision (ICCV)},
  year={2019},
  pages={4947-4956}
}
Saliency methods can aid understanding of deep neural networks. [...] Key Method In this paper, we 1) present a novel region-based attribution method, XRAI, that builds upon integrated gradients (Sundararajan et al. 2017), 2) introduce evaluation methods for empirically assessing the quality of image-based saliency maps (Performance Information Curves (PICs)), and 3) contribute an axiom-based sanity check for attribution methods. Through empirical experiments and example results, we show that XRAI produces…Expand
25 Citations
A-FMI: LEARNING ATTRIBUTIONS FROM DEEP NET-
  • 2020
  • Highly Influenced
There and Back Again: Revisiting Backpropagation Saliency Methods
  • 15
  • PDF
Group-CAM: Group Score-Weighted Visual Explanations for Deep Convolutional Networks
  • Qinglong Zhang, Lu Rao, Yubin Yang
  • Computer Science
  • ArXiv
  • 2021
  • Highly Influenced
  • PDF
Attribution in Scale and Space
  • 8
  • Highly Influenced
  • PDF
Generating Attribution Maps with Disentangled Masked Backpropagation
  • PDF
SWAG: Superpixels Weighted by Average Gradients for Explanations of CNNs
  • 1
  • Highly Influenced
  • PDF
Assessing the validity of saliency maps for abnormality localization in medical imaging
  • 4
  • PDF
Scientific Discovery by Generating Counterfactuals using Image Translation
  • 1
  • PDF
...
1
2
3
...

References

SHOWING 1-10 OF 34 REFERENCES
Sanity Checks for Saliency Maps
  • 437
  • Highly Influential
  • PDF
The (Un)reliability of saliency methods
  • 223
  • PDF
Towards better understanding of gradient-based attribution methods for Deep Neural Networks
  • 293
  • Highly Influential
  • PDF
Visualizing and Understanding Convolutional Networks
  • 9,274
  • PDF
Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
  • 3,187
  • PDF
A Note about: Local Explanation Methods for Deep Neural Networks lack Sensitivity to Parameter Values
  • 8
  • PDF
Real Time Image Saliency for Black Box Classifiers
  • 231
  • Highly Influential
  • PDF
Exploring Principled Visualizations for Deep Network Attributions
  • 8
  • PDF
SmoothGrad: removing noise by adding noise
  • 580
  • PDF
Interpretable Explanations of Black Boxes by Meaningful Perturbation
  • Ruth C. Fong, A. Vedaldi
  • Computer Science, Mathematics
  • 2017 IEEE International Conference on Computer Vision (ICCV)
  • 2017
  • 598
  • PDF
...
1
2
3
4
...