TY - GEN
T1 - Integrating deep features for material recognition
AU - Zhang, Yan
AU - Ozay, Mete
AU - Liu, Xing
AU - Okatani, Takayuki
N1 - Publisher Copyright:
© 2016 IEEE.
Copyright:
Copyright 2019 Elsevier B.V., All rights reserved.
PY - 2016/1/1
Y1 - 2016/1/1
N2 - This paper considers the problem of material recognition. Motivated by observation of close interconnections between material and object recognition, we study how to select and integrate multiple features obtained by different models of Convolutional Neural Networks (CNNs) trained in a transfer learning setting. To be specific, we first compute activations of features using representations on images to select a set of samples which are best represented by the features. Then, we measure uncertainty of the features by computing entropy of class distributions for each sample set. Finally, we compute contribution of each feature to representation of classes for feature selection and integration. Experimental results show that the proposed method achieves state-of-the-art performance on two benchmark datasets for material recognition. Additionally, we introduce a new material dataset, named EFMD, which extends Flickr Material Database (FMD). By the employment of the EFMD for transfer learning, we achieve 84.0% ± 1.8% accuracy on the FMD dataset, which is close to the reported human performance 84.9%.
AB - This paper considers the problem of material recognition. Motivated by observation of close interconnections between material and object recognition, we study how to select and integrate multiple features obtained by different models of Convolutional Neural Networks (CNNs) trained in a transfer learning setting. To be specific, we first compute activations of features using representations on images to select a set of samples which are best represented by the features. Then, we measure uncertainty of the features by computing entropy of class distributions for each sample set. Finally, we compute contribution of each feature to representation of classes for feature selection and integration. Experimental results show that the proposed method achieves state-of-the-art performance on two benchmark datasets for material recognition. Additionally, we introduce a new material dataset, named EFMD, which extends Flickr Material Database (FMD). By the employment of the EFMD for transfer learning, we achieve 84.0% ± 1.8% accuracy on the FMD dataset, which is close to the reported human performance 84.9%.
UR - http://www.scopus.com/inward/record.url?scp=85019080123&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85019080123&partnerID=8YFLogxK
U2 - 10.1109/ICPR.2016.7900209
DO - 10.1109/ICPR.2016.7900209
M3 - Conference contribution
AN - SCOPUS:85019080123
T3 - Proceedings - International Conference on Pattern Recognition
SP - 3697
EP - 3702
BT - 2016 23rd International Conference on Pattern Recognition, ICPR 2016
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 23rd International Conference on Pattern Recognition, ICPR 2016
Y2 - 4 December 2016 through 8 December 2016
ER -