Principal Feature Visualisation in Convolutional Neural Networks
Abstract
We introduce a new visualisation technique for CNNs called Principal Feature Visualisation (PFV). It uses a single forward pass of the original network to map principal features from the final convolutional layer to the original image space as RGB channels. By working on a batch of images we can extract contrasting features, not just the most dominant ones with respect to the classification. This allows us to differentiate between several features in one image in an unsupervised manner. This enables us to assess the feasibility of transfer learning and to debug a pre-trained classifier by localising misleading or missing features.
Cite
Text
Bakken et al. "Principal Feature Visualisation in Convolutional Neural Networks." Proceedings of the European Conference on Computer Vision (ECCV), 2020. doi:10.1007/978-3-030-58592-1_2Markdown
[Bakken et al. "Principal Feature Visualisation in Convolutional Neural Networks." Proceedings of the European Conference on Computer Vision (ECCV), 2020.](https://mlanthology.org/eccv/2020/bakken2020eccv-principal/) doi:10.1007/978-3-030-58592-1_2BibTeX
@inproceedings{bakken2020eccv-principal,
title = {{Principal Feature Visualisation in Convolutional Neural Networks}},
author = {Bakken, Marianne and Kvam, Johannes and Stepanov, Alexey A. and Berge, Asbjørn},
booktitle = {Proceedings of the European Conference on Computer Vision (ECCV)},
year = {2020},
doi = {10.1007/978-3-030-58592-1_2},
url = {https://mlanthology.org/eccv/2020/bakken2020eccv-principal/}
}