Multiterminal Pathfinding in Practical VLSI Systems with Deep Neural Networks

A multiterminal obstacle-avoiding pathfinding approach is proposed. The approach is inspired by deep image learning. The key idea is based on training a conditional generative adversarial network (cGAN) to interpret a pathfinding task as a graphical bitmap and consequently map a pathfinding task onto a pathfinding solution represented by another bitmap. To enable the proposed cGAN pathfinding, a methodology for generating synthetic dataset is also proposed. The cGAN model is implemented in Python/Keras, trained on synthetically generated data, evaluated on practical VLSI benchmarks, and compared with state-of-the-art. Due to effective parallelization on GPU hardware, the proposed approach yields a state-of-the-art like wirelength and a better runtime and throughput for moderately complex pathfinding tasks. However, the runtime and throughput with the proposed approach remain constant with an increasing task complexity, promising orders of magnitude improvement over state-of-the-art in complex pathfinding tasks. The cGAN pathfinder can be exploited in numerous high throughput applications, such as, navigation, tracking, and routing in complex VLSI systems. The last is of particular interest to this work.

Privacy Preserving in Smart Cities Using Various Computing Technologies

IoT has influenced our daily lives through various applications. The high possibility of sensing and publishing sensitive data in the smart environment leads to significant issues: (1) privacy-preserving and (2) real-time services. Privacy is a complex and a subjective notion as its understanding and perception differ among individuals, hence the observation that current studies lack addressing these challenges. This chapter proposes a new privacy-preserving method for IoT devices in the smart city by leveraging ontology, a data model, at the edge of the network. Based on the simulation results using Protege and Visual Studio on a synthetic dataset, the authors find that the solution provides privacy at real-time while addressing heterogeneity issue so that many IoT devices can afford it. Thus, the proposed solution can be widely used for smart cities.

U-Net-Based Foreign Object Detection Method Using Effective Image Acquisition System: A Case of Almond and Green Onion Flake Food Process

Supervised deep learning-based foreign object detection algorithms are tedious, costly, and time-consuming because they usually require a large number of training datasets and annotations. These disadvantages make them frequently unsuitable for food quality evaluation and food manufacturing processes. However, the deep learning-based foreign object detection algorithm is an effective method to overcome the disadvantages of conventional foreign object detection methods mainly used in food inspection. For example, color sorter machines cannot detect foreign objects with a color similar to food, and the performance is easily degraded by changes in illuminance. Therefore, to detect foreign objects, we use a deep learning-based foreign object detection algorithm (model). In this paper, we present a synthetic method to efficiently acquire a training dataset of deep learning that can be used for food quality evaluation and food manufacturing processes. Moreover, we perform data augmentation using color jitter on a synthetic dataset and show that this approach significantly improves the illumination invariance features of the model trained on synthetic datasets. The F1-score of the model that trained the synthetic dataset of almonds at 360 lux illumination intensity achieved a performance of 0.82, similar to the F1-score of the model that trained the real dataset. Moreover, the F1-score of the model trained with the real dataset combined with the synthetic dataset achieved better performance than the model trained with the real dataset in the change of illumination. In addition, compared with the traditional method of using color sorter machines to detect foreign objects, the model trained on the synthetic dataset has obvious advantages in accuracy and efficiency. These results indicate that the synthetic dataset not only competes with the real dataset, but they also complement each other.

QDC-2D: A Semi-Automatic Tool for 2D Analysis of Discontinuities for Rock Mass Characterization

Quantitative characterization of discontinuities is fundamental to define the mechanical behavior of discontinuous rock masses. Several techniques for the semi-automatic and automatic extraction of discontinuities and their properties from raw or processed point clouds have been introduced in the literature to overcome the limits of conventional field surveys and improve data accuracy. However, most of these techniques do not allow characterizing flat or subvertical outcrops because planar surfaces are difficult to detect within point clouds in these circumstances, with the drawback of undersampling the data and providing inappropriate results. In this case, 2D analysis on the fracture traces are more appropriate. Nevertheless, to our knowledge, few methods to perform quantitative analyses on discontinuities from orthorectified photos are publicly available and do not provide a complete characterization. We implemented scanline and window sampling methods in a digital environment to characterize rock masses affected by discontinuities perpendicular to the bedding from trace maps, thus exploiting the potentiality of remote sensing techniques for subvertical and low-relief outcrops. The routine, named QDC-2D (Quantitative Discontinuity Characterization, 2D) was compiled in MATLAB by testing a synthetic dataset and a real case study, from which a high-resolution orthophoto was obtained by means of Structure from Motion technique. Starting from a trace map, the routine semi-automatically classifies the discontinuity sets and calculates their mean spacing, frequency, trace length, and persistence. The fracture network is characterized by means of trace length, intensity, and density estimators. The block volume and shape are also estimated by adding information on the third dimension. The results of the 2D analysis agree with the input used to produce the synthetic dataset and with the data collected in the field by means of conventional geostructural and geomechanical techniques, ensuring the procedure’s reliability. The outcomes of the analysis were implemented in a Discrete Fracture Network model to evaluate their applicability for geomechanical modeling.

Short Circuit Recognition for Metal Electrorefining Using an Improved Faster R-CNN With Synthetic Infrared Images

This paper is concerned with the problem of short circuit detection in infrared image for metal electrorefining with an improved Faster Region-based Convolutional Neural Network (Faster R-CNN). To address the problem of insufficient label data, a framework for automatically generating labeled infrared images is proposed. After discussing factors that affect sample diversity, background, object shape, and gray scale distribution are established as three key variables for synthesis. Raw infrared images without fault are used as backgrounds. By simulating the other two key variables on the background, different classes of objects are synthesized. To improve the detection rate of small scale targets, an attention module is introduced in the network to fuse the semantic segment results of U-Net and the synthetic dataset. In this way, the Faster R-CNN can obtain rich representation ability about small scale object on the infrared images. Strategies of parameter tuning and transfer learning are also applied to improve the detection precision. The detection system trains on only synthetic dataset and tests on actual images. Extensive experiments on different infrared datasets demonstrate the effectiveness of the synthetic methods. The synthetically trained network obtains a mAP of 0.826, and the recall rate of small latent short circuit is superior to that of Faster R-CNN and U-Net, effectively avoiding short-circuit missed detection.

A comprehensive and synthetic dataset for global, regional, and national greenhouse gas emissions by sector 1970–2018 with an extension to 2019

To track progress towards keeping global warming well below 2 ∘C or even 1.5 ∘C, as agreed in the Paris Agreement, comprehensive up-to-date and reliable information on anthropogenic emissions and removals of greenhouse gas (GHG) emissions is required. Here we compile a new synthetic dataset on anthropogenic GHG emissions for 1970–2018 with a fast-track extension to 2019. Our dataset is global in coverage and includes CO2 emissions, CH4 emissions, N2O emissions, as well as those from fluorinated gases (F-gases: HFCs, PFCs, SF6, NF3) and provides country and sector details. We build this dataset from the version 6 release of the Emissions Database for Global Atmospheric Research (EDGAR v6) and three bookkeeping models for CO2 emissions from land use, land-use change, and forestry (LULUCF). We assess the uncertainties of global greenhouse gases at the 90 % confidence interval (5th–95th percentile range) by combining statistical analysis and comparisons of global emissions inventories and top-down atmospheric measurements with an expert judgement informed by the relevant scientific literature. We identify important data gaps for F-gas emissions. The agreement between our bottom-up inventory estimates and top-down atmospheric-based emissions estimates is relatively close for some F-gas species (∼ 10 % or less), but estimates can differ by an order of magnitude or more for others. Our aggregated F-gas estimate is about 10 % lower than top-down estimates in recent years. However, emissions from excluded F-gas species such as chlorofluorocarbons (CFCs) or hydrochlorofluorocarbons (HCFCs) are cumulatively larger than the sum of the reported species. Using global warming potential values with a 100-year time horizon from the Sixth Assessment Report by the Intergovernmental Panel on Climate Change (IPCC), global GHG emissions in 2018 amounted to 58 ± 6.1 GtCO2 eq. consisting of CO2 from fossil fuel combustion and industry (FFI) 38 ± 3.0 GtCO2, CO2-LULUCF 5.7 ± 4.0 GtCO2, CH4 10 ± 3.1 GtCO2 eq., N2O 2.6 ± 1.6 GtCO2 eq., and F-gases 1.3 ± 0.40 GtCO2 eq. Initial estimates suggest further growth of 1.3 GtCO2 eq. in GHG emissions to reach 59 ± 6.6 GtCO2 eq. by 2019. Our analysis of global trends in anthropogenic GHG emissions over the past 5 decades (1970–2018) highlights a pattern of varied but sustained emissions growth. There is high confidence that global anthropogenic GHG emissions have increased every decade, and emissions growth has been persistent across the different (groups of) gases. There is also high confidence that global anthropogenic GHG emissions levels were higher in 2009–2018 than in any previous decade and that GHG emissions levels grew throughout the most recent decade. While the average annual GHG emissions growth rate slowed between 2009 and 2018 (1.2 % yr−1) compared to 2000–2009 (2.4 % yr−1), the absolute increase in average annual GHG emissions by decade was never larger than between 2000–2009 and 2009–2018. Our analysis further reveals that there are no global sectors that show sustained reductions in GHG emissions. There are a number of countries that have reduced GHG emissions over the past decade, but these reductions are comparatively modest and outgrown by much larger emissions growth in some developing countries such as China, India, and Indonesia. There is a need to further develop independent, robust, and timely emissions estimates across all gases. As such, tracking progress in climate policy requires substantial investments in independent GHG emissions accounting and monitoring as well as in national and international statistical infrastructures. The data associated with this article (Minx et al., 2021) can be found at https://doi.org/10.5281/zenodo.5566761.