TY - GEN
T1 - EXTRACTING AND DISTILLING DIRECTION-ADAPTIVE KNOWLEDGE FOR LIGHTWEIGHT OBJECT DETECTION IN REMOTE SENSING IMAGES
AU - Huang, Zhanchao
AU - Li, Wei
AU - Tao, Ran
N1 - Publisher Copyright:
© 2022 IEEE
PY - 2022
Y1 - 2022
N2 - Recently, some lightweight convolutional neural network (CNN) models have been proposed for airborne or spaceborne remote sensing object detection (RSOD) tasks. However, these lightweight detectors suffer from performance degradation due to the compromise of limited computing resources on embedded devices. In order to narrow this performance gap, a direction-adaptive knowledge extraction and distillation (DKED) method is proposed. Specifically, a dynamic directional convolution (DDC) is developed to extract the typical arbitrary-oriented features, and a direction-adaptive knowledge distillation (DKD) strategy is designed for guiding the lightweight model to learn the intrinsic knowledge of the RSOD task from the high-performance model. Experiments on public datasets demonstrate that the proposed method can effectively improve the performance of the lightweight RSOD model without additional inference costs.
AB - Recently, some lightweight convolutional neural network (CNN) models have been proposed for airborne or spaceborne remote sensing object detection (RSOD) tasks. However, these lightweight detectors suffer from performance degradation due to the compromise of limited computing resources on embedded devices. In order to narrow this performance gap, a direction-adaptive knowledge extraction and distillation (DKED) method is proposed. Specifically, a dynamic directional convolution (DDC) is developed to extract the typical arbitrary-oriented features, and a direction-adaptive knowledge distillation (DKD) strategy is designed for guiding the lightweight model to learn the intrinsic knowledge of the RSOD task from the high-performance model. Experiments on public datasets demonstrate that the proposed method can effectively improve the performance of the lightweight RSOD model without additional inference costs.
KW - Dynamic directional convolution
KW - knowledge distillation
KW - lightweight object detection
KW - remote sensing
UR - http://www.scopus.com/inward/record.url?scp=85134048957&partnerID=8YFLogxK
U2 - 10.1109/ICASSP43922.2022.9747798
DO - 10.1109/ICASSP43922.2022.9747798
M3 - Conference contribution
AN - SCOPUS:85134048957
T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
SP - 2125
EP - 2129
BT - 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 47th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022
Y2 - 23 May 2022 through 27 May 2022
ER -