@inproceedings{mlc_gnn_cvpr17,
author = {Kenneth Marino and
Ruslan Salakhutdinov and
Abhinav Gupta},
title = {The More You Know: Using Knowledge Graphs for Image Classification},
booktitle = {The {IEEE} Conference on Computer Vision and Pattern Recognition (CVPR)},
pages = {20--28},
year = {2017}
}
links
summary
- improving performance on multi-label image classification using structured prior knowledge forms the main goal
- the authors build on GGNNs in order to incorporate large knowledge graphs into a vision classification pipeline
- in experiments, the method outperfroms standard neural net baselines both quantitatively and qualitatively