TY - GEN
T1 - Multi-depth dilated network for fashion landmark detection
AU - Kai, Zeng
AU - Feng, Jun
AU - Sutcliffe, Richard
AU - Xiaoyu, Wang
AU - Qirong, Bu
N1 - Publisher Copyright:
© 2019 IEEE.
PY - 2019/7
Y1 - 2019/7
N2 - The topic of fashion landmark detection has become popular recently, especially with the development of convolutional neural networks. However, there are still some challenges that need to be addressed, including the detection of hard keypoints such as those which are occluded or invisible. To tackle this problem, we propose in this paper a novel Multi-Depth Dilated (MDD) block which is composed of different numbers of dilated convolutions in parallel, so that the MDD can effectively extract large-scale context information at different levels and hence obtain relative positional relationship information between keypoints, information which is necessary for the inference of hard keypoints. Furthermore, we use an Online Hard Keypoints Mining method for training to further boost the effectiveness of hard keypoints detection. Through the experiment, we demonstrate that by stacking the MDD blocks, we construct a Multi-Depth Dilated Network (MDDNet) that achieves state-of-the-art results on fashion benchmark datasets.
AB - The topic of fashion landmark detection has become popular recently, especially with the development of convolutional neural networks. However, there are still some challenges that need to be addressed, including the detection of hard keypoints such as those which are occluded or invisible. To tackle this problem, we propose in this paper a novel Multi-Depth Dilated (MDD) block which is composed of different numbers of dilated convolutions in parallel, so that the MDD can effectively extract large-scale context information at different levels and hence obtain relative positional relationship information between keypoints, information which is necessary for the inference of hard keypoints. Furthermore, we use an Online Hard Keypoints Mining method for training to further boost the effectiveness of hard keypoints detection. Through the experiment, we demonstrate that by stacking the MDD blocks, we construct a Multi-Depth Dilated Network (MDDNet) that achieves state-of-the-art results on fashion benchmark datasets.
KW - Convolutional Neural Network
KW - Deep Learning
KW - Fashion Landmark Detection
UR - http://www.scopus.com/inward/record.url?scp=85071499300&partnerID=8YFLogxK
U2 - 10.1109/ICMEW.2019.00018
DO - 10.1109/ICMEW.2019.00018
M3 - Conference contribution
AN - SCOPUS:85071499300
T3 - Proceedings - 2019 IEEE International Conference on Multimedia and Expo Workshops, ICMEW 2019
SP - 60
EP - 65
BT - Proceedings - 2019 IEEE International Conference on Multimedia and Expo Workshops, ICMEW 2019
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2019 IEEE International Conference on Multimedia and Expo Workshops, ICMEW 2019
Y2 - 8 July 2019 through 12 July 2019
ER -