TY - GEN
T1 - Label Smoothing Technique for Ordinal Classification in Cloud Assessment
AU - Wei, Yuxuan
AU - Liu, Qixuan
AU - Zhang, Guixu
AU - Peng, Yaxin
AU - Shen, Chaomin
N1 - Publisher Copyright:
© 2020 IEEE.
PY - 2020/9/26
Y1 - 2020/9/26
N2 - Satellite image classification is a challenging task if the input labels are not sufficiently accurate. The automatic cloud cover assessment (ACCA), for example, aims to classify the cloud covers of satellite images as alphabetical categories from A to E showing the escalating levels of clouds; however, those labels for training are often obtained by a subjective qualitative assessment, i.e., they may be not accurate. Therefore, this paper studies how to conduct ACCA under this circumstance. We propose a label smoothing approach and improve the accuracy around 3 percentage points (e.g., from 75.9% to 78.4% for ResNet network) without changing other network structures and parameters.
AB - Satellite image classification is a challenging task if the input labels are not sufficiently accurate. The automatic cloud cover assessment (ACCA), for example, aims to classify the cloud covers of satellite images as alphabetical categories from A to E showing the escalating levels of clouds; however, those labels for training are often obtained by a subjective qualitative assessment, i.e., they may be not accurate. Therefore, this paper studies how to conduct ACCA under this circumstance. We propose a label smoothing approach and improve the accuracy around 3 percentage points (e.g., from 75.9% to 78.4% for ResNet network) without changing other network structures and parameters.
KW - Ordinary classification
KW - cloud cover assessment
KW - label smoothing
KW - loss function
KW - neural network
UR - https://www.scopus.com/pages/publications/85101970578
U2 - 10.1109/IGARSS39084.2020.9323714
DO - 10.1109/IGARSS39084.2020.9323714
M3 - 会议稿件
AN - SCOPUS:85101970578
T3 - International Geoscience and Remote Sensing Symposium (IGARSS)
SP - 2264
EP - 2267
BT - 2020 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2020 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2020 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2020
Y2 - 26 September 2020 through 2 October 2020
ER -