Weakly Supervised Segmentation of Histopathology Images: An Insight in Feature Maps Ability for Learning Models Interpretation

Yanbo Feng, Adel Hafiane, Hélène Laurent

2022

Abstract

Feature map is obtained from the middle layer of convolutional neural network (CNN), it carries the regional information captured by network itself about the target of input image. This property is widely used in weakly supervised learning to achieve target localization and segmentation. However, the traditional method of processing feature map is often associated with the weight of output layer. In this paper, the weak correlation between feature map and weight is discussed. We believe that it is not accurate to directly transplant the weights of output layer to feature maps, the reason is that the global mean value of feature map loses its spatial information, weighting scalars cannot accurately constrain the three-dimensional feature maps. We highlight that the feature map in a specific channel has invariance to target’s location, it can stably activate the more complete region directly related to target, that is, the feature map ability has strong correlation with the channel.

Download


Paper Citation


in Harvard Style

Feng Y., Hafiane A. and Laurent H. (2022). Weakly Supervised Segmentation of Histopathology Images: An Insight in Feature Maps Ability for Learning Models Interpretation. In Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 5: VISAPP; ISBN 978-989-758-555-5, SciTePress, pages 421-427. DOI: 10.5220/0010830400003124


in Bibtex Style

@conference{visapp22,
author={Yanbo Feng and Adel Hafiane and Hélène Laurent},
title={Weakly Supervised Segmentation of Histopathology Images: An Insight in Feature Maps Ability for Learning Models Interpretation},
booktitle={Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 5: VISAPP},
year={2022},
pages={421-427},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010830400003124},
isbn={978-989-758-555-5},
}


in EndNote Style

TY - CONF

JO - Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 5: VISAPP
TI - Weakly Supervised Segmentation of Histopathology Images: An Insight in Feature Maps Ability for Learning Models Interpretation
SN - 978-989-758-555-5
AU - Feng Y.
AU - Hafiane A.
AU - Laurent H.
PY - 2022
SP - 421
EP - 427
DO - 10.5220/0010830400003124
PB - SciTePress