Bounding overwatch

About: Bounding overwatch is a research topic. Over the lifetime, 966 publications have been published within this topic receiving 15156 citations.

GLE-Net: A Global and Local Ensemble Network for Aerial Object Detection

Abstract: Abstract Recent advances in camera-equipped drone applications increased the demand for visual object detection algorithms with deep learning for aerial images. There are several limitations in accuracy for a single deep learning model. Inspired by ensemble learning can significantly improve the generalization ability of the model in the machine learning field, we introduce a novel integration strategy to combine the inference results of two different methods without non-maximum suppression. In this paper, a global and local ensemble network (GLE-Net) was proposed to increase the quality of predictions by considering the global weights for different models and adjusting the local weights for bounding boxes. Specifically, the global module assigns different weights to models. In the local module, we group the bounding boxes that corresponding to the same object as a cluster. Each cluster generates a final predict box and assigns the highest score in the cluster as the score of the final predict box. Experiments on benchmarks VisDrone2019 show promising performance of GLE-Net compared with the baseline network.

State bounding estimation for a linear continuous-time singular system with time-varying delay

Abstract: This paper investigates the problem of a state bounding estimation for a linear continuous-time singular system with time-varying delay. By employing the maximal Lyapunov–Krasovskii functional and applying the new free-matrix-based integral inequality, some proper conditions are derived in terms of LMIs and a bounding estimation lemma and set are obtained for the studied singular system.

Post-OCR Paragraph Recognition by Graph Convolutional Networks

Abstract: We propose a new approach for paragraph recognition in document images by spatial graph convolutional networks (GCN) applied on OCR text boxes. Two steps, namely line splitting and line clustering, are performed to extract paragraphs from the lines in OCR results. Each step uses a β-skeleton graph constructed from bounding boxes, where the graph edges provide efficient support for graph convolution operations. With pure layout input features, the GCN model size is 3~4 orders of magnitude smaller compared to RCNN based models, while achieving comparable or better accuracies on PubLayNet and other datasets. Furthermore, the GCN models show good generalization from synthetic training data to real-world images, and good adaptivity for variable document styles.

Bounding Techniques for Model-Structure Selection

Abstract: Structure-selection methods for regression-type models are discussed. Selection is based on parameter bounds, obtained by plotting in parameter space known deterministic bounds on the error between model output and observed output. One method examines the effect of imposing bounds on the sample autocorrelation of the errors, another detects near-degeneracy of the parameter bounds, and two methods aim to delete whichever parameter or combination of parameters is least well defined by the observations. The last two methods have close parallels with principal-component analysis and singular-value decomposition.

Range bounding with Taylor models: some case studies

Abstract: Taylor model methods represent a combination of high-order multivariate automatic differentiation and the simultaneous computation of an interval remainder bound enclosing approximation error over a given domain. This method allows a far-reaching suppression of the dependency problem common to interval methods, and can thus often be used for precise range bounding problems. We compare the performance of the method with other state of the art validated tools including centered forms and mean value forms. We also compare with the computation of remainder bounds via high-order interval automatic differentiation.