TY - GEN
T1 - Efficient relative attribute learning using graph neural networks
AU - Meng, Zihang
AU - Adluru, Nagesh
AU - Kim, Hyunwoo J.
AU - Fung, Glenn
AU - Singh, Vikas
N1 - Funding Information:
Acknowledgment. This work was partially supported by funding from American Family Insurance and UW CPCP AI117924. Partial support from NSF CAREER award 1252725, NIH grants R01 AG040396, BRAIN Initiative R01-EB022883, and Waisman IDDRC U54-HD090256 is also acknowledged. The authors are grateful to Haoliang Sun for help with illustrations and other suggestions/advice on this project. The code will appear in https://github.com/zihangm/RAL GNN.
Publisher Copyright:
© 2018, Springer Nature Switzerland AG.
PY - 2018
Y1 - 2018
N2 - A sizable body of work on relative attributes provides evidence that relating pairs of images along a continuum of strength pertaining to a visual attribute yields improvements in a variety of vision tasks. In this paper, we show how emerging ideas in graph neural networks can yield a solution to various problems that broadly fall under relative attribute learning. Our main idea is the observation that relative attribute learning naturally benefits from exploiting the graph of dependencies among the different relative attributes of images, especially when only partial ordering is provided at training time. We use message passing to perform end to end learning of the image representations, their relationships as well as the interplay between different attributes. Our experiments show that this simple framework is effective in achieving competitive accuracy with specialized methods for both relative attribute learning and binary attribute prediction, while relaxing the requirements on the training data and/or the number of parameters, or both.
AB - A sizable body of work on relative attributes provides evidence that relating pairs of images along a continuum of strength pertaining to a visual attribute yields improvements in a variety of vision tasks. In this paper, we show how emerging ideas in graph neural networks can yield a solution to various problems that broadly fall under relative attribute learning. Our main idea is the observation that relative attribute learning naturally benefits from exploiting the graph of dependencies among the different relative attributes of images, especially when only partial ordering is provided at training time. We use message passing to perform end to end learning of the image representations, their relationships as well as the interplay between different attributes. Our experiments show that this simple framework is effective in achieving competitive accuracy with specialized methods for both relative attribute learning and binary attribute prediction, while relaxing the requirements on the training data and/or the number of parameters, or both.
KW - Graph neural networks
KW - Message passing
KW - Multi-task learning
KW - Relative attribute learning
UR - http://www.scopus.com/inward/record.url?scp=85055714853&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-01264-9_34
DO - 10.1007/978-3-030-01264-9_34
M3 - Conference contribution
AN - SCOPUS:85055714853
SN - 9783030012632
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 575
EP - 590
BT - Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings
A2 - Ferrari, Vittorio
A2 - Sminchisescu, Cristian
A2 - Weiss, Yair
A2 - Hebert, Martial
PB - Springer Verlag
T2 - 15th European Conference on Computer Vision, ECCV 2018
Y2 - 8 September 2018 through 14 September 2018
ER -