TY - GEN
T1 - A simple rotational equivariance loss for generic convolutional segmentation networks
T2 - 39th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2019
AU - Lin, Kangcheng
AU - Huang, Bohao
AU - Collins, Leslie M.
AU - Bradbury, Kyle
AU - Malof, Jordan M.
N1 - Publisher Copyright:
© 2019 IEEE.
PY - 2019/7
Y1 - 2019/7
N2 - Segmentation convolutional neural networks (SCNNs) are now popular for the semantic segmentation (i.e., dense pixel-wise labeling) of remote sensing imagery, such as color or hyperspectral satellite imagery. One desirable property of SCNNs when applied to remote sensing problems is rotational equivariance. This property implies that the class label assigned to a particular pixel (building, road, etc.) does not change if the input imagery is rotated by an arbitrary angle. We argue that recently proposed methods to make rotational equivariant SCNNs fall into two broad categories: easily employed methods that are somewhat ineffective, and highly effective methods that are complicated and potentially incompatible with state-of-the-art SCNN techniques. We propose a simple addition to the standard SCNN loss function that encourages the SCNN to be rotationally equivariant, and is easily added to modern SCNNs. We test the method on the Inria building labeling dataset and compare it to the popular simple approach of adding random rotational augmentations of the input imagery during training. We show that the proposed approach (i) achieves improved equivariance and (ii) yields performance improvements on average.
AB - Segmentation convolutional neural networks (SCNNs) are now popular for the semantic segmentation (i.e., dense pixel-wise labeling) of remote sensing imagery, such as color or hyperspectral satellite imagery. One desirable property of SCNNs when applied to remote sensing problems is rotational equivariance. This property implies that the class label assigned to a particular pixel (building, road, etc.) does not change if the input imagery is rotated by an arbitrary angle. We argue that recently proposed methods to make rotational equivariant SCNNs fall into two broad categories: easily employed methods that are somewhat ineffective, and highly effective methods that are complicated and potentially incompatible with state-of-the-art SCNN techniques. We propose a simple addition to the standard SCNN loss function that encourages the SCNN to be rotationally equivariant, and is easily added to modern SCNNs. We test the method on the Inria building labeling dataset and compare it to the popular simple approach of adding random rotational augmentations of the input imagery during training. We show that the proposed approach (i) achieves improved equivariance and (ii) yields performance improvements on average.
KW - aerial imagery
KW - building detection
KW - convolutional neural networks
KW - deep learning
KW - semantic segmentation
UR - http://www.scopus.com/inward/record.url?scp=85077678377&partnerID=8YFLogxK
U2 - 10.1109/IGARSS.2019.8898722
DO - 10.1109/IGARSS.2019.8898722
M3 - Conference contribution
AN - SCOPUS:85077678377
T3 - International Geoscience and Remote Sensing Symposium (IGARSS)
SP - 3876
EP - 3879
BT - 2019 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2019 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 28 July 2019 through 2 August 2019
ER -