Recent increases in aerial image access and volume, increases in computational power, and interest in applications have opened the door to scaling up object detection and domain adaptation research to production. Aerial data sets are very large in size, and each frame of the data set contains a huge number of dense and small objects. Deep learning applications for aerial imagery are behind due to a lack of training data, and researchers have recently turned to domain adaptation (DA) from a labeled data set to an unlabeled data set to alleviate the issue. These factors create two major challenges: the high variety between datasets (e.g. object sizes, class distributions, object feature uniformity, image acquisition, distance, weather conditions), and the size of objects in satellite imagery and subsequent failure of state-of-the-art to capture small objects, local features, and region proposals for densely overlapped objects in satellite image. In this paper, we propose two solutions to these problems: a domain discriminator to better align the local feature space between domains; and a novel pipeline that improves the back-end by spatial pyramid pooling, cross-stage partial network, region proposal network via heatmap-based region proposals, and object localization and identification through a novel image difficulty score that adapts the overall focal loss measure based on the image difficulty. Our proposed model outperformed the state-of-the-art method by 7.4%.
翻译:最近航空图像访问量和数量增加,计算能力增加,对应用的兴趣增加,为扩大物体探测和对生产进行领域适应研究打开了大门。空中数据集规模很大,每个数据集框架都包含大量密集和小天体。空中图像的深学习应用落后于培训数据,研究人员最近转向域适应(DA),从标签数据集转向未贴标签数据集,转向缓解问题的未贴标签数据集。这些因素造成了两大挑战:数据集差异很大(例如,物体大小、类分布、物体特征统一、图像获取、距离、天气条件),卫星图像中物体的大小以及随后无法捕捉小物体、地方特征和卫星图像中密集重叠物体的区域提议。在本文件中,我们提出了解决这些问题的两种解决办法:一个域歧视者,以更好地将各区域之间的本地地物空间空间空间空间空间空间空间空间定位;跨级部分网络;区域建议网络,通过基于热映的图像定位系统,通过基于区域测位模型的模型,调整我们基于总体图像测算方法的模型,从而改进了我们地区图像测算的难度。