Electric power operation is necessary for the development of grid companies, where safety monitoring electric difficult. Irregular deformable objects commonly used in electrical construction, such as belts and seines, have a dynamic geometric appearance which leads to poor performance traditional detection methods. This paper proposes an end-to-end instance segmentation method using multi-insta...

Adversarial training is an effective way to defend deep neural networks (DNN) against adversarial examples. However, there are atypical samples that rare and hard learn, or even hurt DNNs' generalization performance on test data. In this paper, we propose a novel algorithm reweight the based self-supervised techniques mitigate negative effects of samples. Specifically, memory bank built record ...

Ship instance segmentation in synthetic aperture radar (SAR) images can provide more detailed location information and shape information, which is of great significance for port ship scheduling traffic management. However, there little research work on SAR segmentation, the general accuracy low because characteristics target task, such as multi-scale, aspect ratio, noise interference, are not c...

In remote sensing, it is often challenging to acquire or collect a large data set that accurately labeled. This difficulty usually due several issues, including but not limited the study site’s spatial area and accessibility, errors in global positioning system (GPS), mixed pixels caused by an image’s resolution. We propose approach, with two variations, estimates multiple-target signatures fro...

In this paper we propose a new algorithm, named NICE, to generate counterfactual explanations for tabular data that specifically takes into account algorithmic requirements often emerge in real-life deployments: (1) the ability provide an explanation all predictions, (2) being able handle any classification model (also non-differentiable ones), (3) efficient run time, and (4) providing multiple...