https://github.com/boschresearch/GridSaliency-ToyDatasetGen

Code for toy dataset generation of "Grid Saliency for Context Explanations of Semantic Segmentation" (https://arxiv.org/abs/1907.13054)
https://github.com/boschresearch/GridSaliency-ToyDatasetGen

bcai paper-resource

Last synced: 12 days ago
JSON representation

Code for toy dataset generation of "Grid Saliency for Context Explanations of Semantic Segmentation" (https://arxiv.org/abs/1907.13054)

• Host: GitHub
• URL: https://github.com/boschresearch/GridSaliency-ToyDatasetGen
• Owner: boschresearch
• License: agpl-3.0
• Archived: true
• Created: 2019-09-20T11:25:31.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2019-11-28T14:09:57.000Z (almost 5 years ago)
• Last Synced: 2024-08-01T20:36:42.161Z (3 months ago)
• Topics: bcai, paper-resource
• Language: Python
• Homepage:
• Size: 4.64 MB
• Stars: 11
• Watchers: 5
• Forks: 1
• Open Issues: 3
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-computer-vision-papers - 2019NIPS

README

        
# GridSaliency-ToyDatasetGen

This repository provides code to generate the toy datasets used in "Grid Saliency for Context Explanations of Semantic Segmentation" ([arxiv](https://arxiv.org/abs/1907.13054), [poster](./poster/poster_grid-saliency.pdf)).

Abstract: Recently, there has been a growing interest in developing saliency methods that provide visual explanations of network predictions. Still, the usability of existing methods is limited to image classification models. To overcome this limitation, we extend the existing approaches to generate grid saliencies, which provide spatially coherent visual explanations for (pixel-level) dense prediction networks. As the proposed grid saliency allows to spatially disentangle the object and its context, we specifically explore its potential to produce context explanations for semantic segmentation networks, discovering which context most influences the class predictions inside a target object area. We investigate the effectiveness of grid saliency on a synthetic dataset with an artificially induced bias between objects and their context as well as on the real-world Cityscapes dataset using state-of-the-art segmentation networks. Our results show that grid saliency can be successfully used to provide easily interpretable context explanations and, moreover, can be employed for detecting and localizing contextual biases present in the data.