文献类型：期刊文献

英文题名：Efficient Hybrid Supervision for Instance Segmentation in Aerial Images

作者：Chen, Linwei[1];Fu, Ying[1];You, Shaodi[2];Liu, Hongzhe[3]

第一作者：Chen, Linwei

通讯作者：Fu, Y[1]

机构：[1]Beijing Inst Technol, Sch Comp Sci & Technol, Beijing 100081, Peoples R China;[2]Univ Amsterdam, Inst Informat, Comp Vis Res Grp, NL-1098 XH Amsterdam, Netherlands;[3]Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing 100101, Peoples R China

第一机构：Beijing Inst Technol, Sch Comp Sci & Technol, Beijing 100081, Peoples R China

通讯机构：[1]corresponding author), Beijing Inst Technol, Sch Comp Sci & Technol, Beijing 100081, Peoples R China.

年份：2021

卷号：13

期号：2

起止页码：1-22

外文期刊名：REMOTE SENSING

收录：;EI(收录号：20210309793045);Scopus(收录号：2-s2.0-85099410275);WOS:【SCI-EXPANDED(收录号:WOS:000611563100001)】；

基金：This research was funded by the National Natural Science Foundation of China under Grants No. 61936011, No. 61871039, No. 62006101, and the Collaborative Innovation Center for Visual Intelligence under Grant No. CYXC2011.

语种：英文

外文关键词：hybrid supervision; instance segmentation; aerial images; labeling cost

摘要：Instance segmentation in aerial images is of great significance for remote sensing applications, and it is inherently more challenging because of cluttered background, extremely dense and small objects, and objects with arbitrary orientations. Besides, current mainstream CNN-based methods often suffer from the trade-off between labeling cost and performance. To address these problems, we present a pipeline of hybrid supervision. In the pipeline, we design an ancillary segmentation model with the bounding box attention module and bounding box filter module. It is able to generate accurate pseudo pixel-wise labels from real-world aerial images for training any instance segmentation models. Specifically, bounding box attention module can effectively suppress the noise in cluttered background and improve the capability of segmenting small objects. Bounding box filter module works as a filter which removes the false positives caused by cluttered background and densely distributed objects. Our ancillary segmentation model can locate object pixel-wisely instead of relying on horizontal bounding box prediction, which has better adaptability to arbitrary oriented objects. Furthermore, oriented bounding box labels are utilized for handling arbitrary oriented objects. Experiments on iSAID dataset show that the proposed method can achieve comparable performance (32.1 AP) to fully supervised methods (33.9 AP), which is obviously higher than weakly supervised setting (26.5 AP), when using only 10% pixel-wise labels.