The demand for the sensor-based detection of camouflage objects widely exists in biological research, remote sensing, and military applications. However, the performance of traditional object detection algorithms is limited, as they are incapable of extracting informative parts from low signal-to-noise ratio features. To address this problem, we propose Camouflaged Object Detection with Cascade and Feedback Fusion (CODCEF), a deep learning framework based on an RGB optical sensor that leverages a cascaded structure with Feedback Partial Decoders (FPD) instead of a traditional encoder–decoder structure. Through a selective fusion strategy and feedback loop, FPD reduces the loss of information and the interference of noises in the process of feature interweaving. Furthermore, we introduce Pixel Perception Fusion (PPF) loss, which aims to pay more attention to local pixels that might become the edges of an object. Experimental results on an edge device show that CODCEF achieved competitive results compared with 10 state-of-the-art methods.
This article analyzes the dynamics of turnout and the political impact of five cycles of protest, consisting of forty-two mass demonstrations that occurred on Mondays in Leipzig over the period 1989–91. These demonstrations are interpreted as an informational cascade that publicly revealed some of the previously hidden information about the malign nature of the East German communist regime. Once this information became publicly available, the viability of the regime was undermined. The Monday demonstrations subsequently died a slow death as their informational role declined.
ABSTRACT:Future projections of coastal erosion, which are one of the most demanded climate services in coastal areas, are mainly developed using top-down approaches. These approaches consist of undertaking a sequence of steps that include selecting emission or concentration scenarios and climate models, correcting models bias, applying downscaling methods, and implementing coastal erosion models. The information involved in this modelling chain cascades across steps, and so does related uncertainty, which accumulates in the results. Here, we develop long-term multi-ensemble probabilistic coastal erosion projections following the steps of the top-down approach, factorise, decompose and visualise the uncertainty cascade using real data and analyse the contribution of the uncertainty sources (knowledge-based and intrinsic) to the total uncertainty. We find a multi-modal response in long-term erosion estimates and demonstrate that not sampling internal climate variability?s uncertainty sufficiently could lead to a truncated outcomes range, affecting decision-making. Additionally, the noise arising from internal variability (rare outcomes) appears to be an important part of the full range of results, as it turns out that the most extreme shoreline retreat events occur for the simulated chronologies of climate forcing conditions. We conclude that, to capture the full uncertainty, all sources need to be properly sampled considering the climate-related forcing variables involved, the degree of anthropogenic impact and time horizon targeted. ; AT acknowledges the financial support from the FENIX Project by the Government of Cantabria. This research was also funded by the Spanish Government via the grant RISKCOADAPT (BIA2017-89401-R).
AbstractDigitally networked and new, unconventional activities allow citizens to participate politically in activities that are low in the effort and risks they bear. At the same time, low-effort types of participation are more loosely connected to democratic political systems, thereby challenging established modes of political decision-making. This can set in motion two competing dynamics: While some citizens move closer to the political system in their activities (upstream effects), others engage in political activities more distant from it (downstream effects). This study investigates non-electoral participation trajectories and tests intra-individual change in political participation types over time, exploring whether such dynamics depend on citizens' exposure to political information. Utilizing a three-wave panel survey (n = 3490) and random intercept cross-lagged panel models with SEM, we find more evidence for downstream effects but detect overall diverse participation trajectories over time and a potentially crucial role of elections for non-electoral participation trajectories.
Automated detection of the content of images remains a challenging problem in artificial intelligence. Hence, continuous manual monitoring of restricted development zones is critical to maintaining territorial integrity and national security. In this regard, local governments of the Republic of Korea conduct four periodic inspections per year to preserve national territories from illegal encroachments and unauthorized developments in restricted zones. The considerable expense makes responding to illegal developments difficult for local governments. To address this challenge, we propose a deep-learning-based Cascade Mask region-based convolutional neural network (R-CNN) algorithm designed to perform automated detection of greenhouses in aerial photographs for efficient and continuous monitoring of restricted development zones in the Republic of Korea. Our proposed model is regional-based because it was optimized for the Republic of Korea via transfer learning and hyperparameter tuning, which improved the efficiency of the automated detection of greenhouse facilities. The experimental results demonstrated that the mAP value of the proposed Cascade Mask R-CNN model was 83.6, which was 12.83 higher than baseline mask R-CNN, and 0.9 higher than Mask R-CNN with hyperparameter tuning and transfer learning considered. Similarly, the F1-score of the proposed Cascade Mask R-CNN model was 62.07, which outperformed those of the baseline mask R-CNN and the Mask R-CNN with hyperparameter tuning and transfer learning considered (i.e., the F1-score 52.33 and 59.13, respectively). The proposed improved Cascade Mask R-CNN model is expected to facilitate efficient and continuous monitoring of restricted development zones through routine screening procedures. Moreover, this work provides a baseline for developing an integrated management system for national-scale land-use planning and development infrastructure by synergizing geographical information systems, remote sensing, and deep learning models.
Automated detection of the content of images remains a challenging problem in artificial intelligence. Hence, continuous manual monitoring of restricted development zones is critical to maintaining territorial integrity and national security. In this regard, local governments of the Republic of Korea conduct four periodic inspections per year to preserve national territories from illegal encroachments and unauthorized developments in restricted zones. The considerable expense makes responding to illegal developments difficult for local governments. To address this challenge, we propose a deep-learning-based Cascade Mask region-based convolutional neural network (R-CNN) algorithm designed to perform automated detection of greenhouses in aerial photographs for efficient and continuous monitoring of restricted development zones in the Republic of Korea. Our proposed model is regional-based because it was optimized for the Republic of Korea via transfer learning and hyperparameter tuning, which improved the efficiency of the automated detection of greenhouse facilities. The experimental results demonstrated that the mAP value of the proposed Cascade Mask R-CNN model was 83.6, which was 12.83 higher than baseline mask R-CNN, and 0.9 higher than Mask R-CNN with hyperparameter tuning and transfer learning considered. Similarly, the F1-score of the proposed Cascade Mask R-CNN model was 62.07, which outperformed those of the baseline mask R-CNN and the Mask R-CNN with hyperparameter tuning and transfer learning considered (i.e., the F1-score 52.33 and 59.13, respectively). The proposed improved Cascade Mask R-CNN model is expected to facilitate efficient and continuous monitoring of restricted development zones through routine screening procedures. Moreover, this work provides a baseline for developing an integrated management system for national-scale land-use planning and development infrastructure by synergizing geographical information systems, remote sensing, and deep learning models.
Interstate 90 over the Cascades is a significant barrier to over 250 species of wildlife, including cougar, elk, deer, mustelids (otters, fishers, badgers, etc.), amphibians, and reptiles. In the vicinity of Snoqualmie Pass, urban development to the west and agriculture and resort development on the east has shrunk the forest connecting the north and south Cascades to less than 64.6 kilometers wide. The Washington State Department of Transportation (WSDOT) is proposing to expand a 24.15-kilometer stretch of Interstate 90 just east of Snoqualmie Pass through a particularly critical zone for north-south wildlife corridors. Absent effective wildlife-crossing structures, the expansion would worsen the barrier by increasing roadkill and further isolating populations, thus inhibiting genetic exchange. However, the state has made ecological connectivity a project goal, along with increasing capacity, straightening curves, and repaving. The I-90 Wildlife Bridges Coalition has been working with WSDOT, other public officials, transportation interests, and the public to promote high-quality wildlife-crossing structures. Such structures can also improve safety for motorists by reducing collisions that are sometimes fatal to humans, as well as wildlife. Good data is available to inform where to build crossing structures. WSDOT and the US Forest Service collaborated on a study entitled I-90 Snoqualmie Pass Wildlife Habitat Linkage Assessment (Singleton and Lehmkuhl 2000) that used tracking and road-kill counts to map existing crossing activity. Additional relevant information comes from analysis leading to the Snoqualmie Pass Adaptive Management Area Plan and I-90 Land Exchange (US Forest Service, 1997 and 1999) and Washington State Dept. of Fish and Wildlife studies of cougar movements using radio collars. Recent land acquisitions and national forest-management changes have dramatically improved the outlook for habitat quality near the project. In recent years, purchases, donations, and exchanges have brought more than 50,000 acres of land valued at $200 million into public ownership and protection. The Forest Service is committing to additional habitat restoration, such as road removal. Two of the distinguishing features of the I-90 project are the prevalence of wetlands associated with the Yakima River and the variation in habitat as precipitation and elevation decline from west to east. A variety of structure types—from extended vehicle bridges, to box culverts, to overpasses specifically for wildlife—is required to allow both hydrological connectivity and connections for a diverse array of species. Preferred habitat conditions and existing movement patterns are balanced with site-specific design considerations, including cost, to establish a range of possible solutions to be presented in a draft environmental-impact statement due in spring 2005. Given the intense competition for transportation funds, particularly big-ticket projects near urban areas, the I-90 Snoqualmie Pass East project will need broad-based support to obtain funding. To overcome the environmental community's general opposition to expanded freeways, the project will need to provide a high level of wildlife connectivity. Project proponents will also need to navigate anti-tax politics by joining in a diverse coalition of agencies, conservation groups, and shipping interests. The recent partnership to acquire habitat north and south of the project points the way. The coalition has grown out of a history of grassroots activism and collaboration around the Central Cascades region. Citizen involvement has played a critical role in the management policies of this area. The I-90 project will be a greater success due to the high level of attention and input received from the public. Public involvement will have peaked in the spring of 2005 with the release of the Draft Environmental Impact Statement followed by five public comment hearings throughout Washington State. This input will be considered throughout the summer of 2005 and (hopefully) brought to a successful completion in the fall/winter of the same year.
Intracavity Laser Absorption Spectroscopy (ICLAS) at IR wavelengths offers an opportunity for spectral sensing of low vapor pressure compounds. We report here an ICLAS system design based on a quantum cascade laser (QCL) at THz (69.9 ?m) and IR wavelengths (9.38 and 8.1 ?m) with an open external cavity. The sensitivity of such a system is potentially very high due to extraordinarily long effective optical paths that can be achieved in an active cavity. Sensitivity estimation by numerical solution of the laser rate equations for the THz QCL ICLAS system is determined. Experimental development of the external cavity QCL is demonstrated for the two IR wavelengths, as supported by appearance of fine mode structure in the laser spectrum. The 8.1 ?m wavelength exhibits a dramatic change in the output spectrum caused by the weak intracavity absorption of acetone. Numerical solution of the laser rate equations yields a sensitivity estimation of acetone partial pressure of 165 mTorr corresponding to ~ 200 ppm. The system is also found sensitive to the humidity in the laboratory air with an absorption coefficient of just 3 x 10-7 cm-1 indicating a sensitivity of 111 ppm. Reported also is the design of a compact integrated data acquisition and control system. Potential applications include military and commercial sensing for threat compounds such as explosives, chemical gases, biological aerosols, drugs, banned or invasive organisms, bio-medical breath analysis, and terrestrial or planetary atmospheric science. ; 2011-12-01 ; Ph.D. ; Sciences, Physics ; Doctoral ; This record was generated from author submitted information.
Despite the intensive study of the viral spread of fake news in political echo chambers (ECs) on social networking services (SNSs), little is known regarding the underlying structure of the daily information spread in these ECs. Moreover, the effect of SNSs on opinion polarisation is still unclear in terms of pluralistic information access or selective exposure to opinions in an SNS. In this study, we confirmed the steady, highly independent nature of left- and right-leaning ECs, both of which are composed of approximately 250,000 users, from a year-long reply/retweet network of 42 million Japanese Twitter users. We found that both communities have similarly efficient information spreading networks with densely connected and core-periphery structures. Core nodes resonate in the early stages of information cascades, and unilaterally transmit information to peripheral nodes. Each EC has resonant core users who amplify and steadily spread information to a quarter of a million users. In addition, we confirmed the existence of extremely aggressive users of ECs who co-reply/retweet each other. The connection between these users and top influencers suggests that the extreme opinions of the former group affect the entire community through the top influencers.
chapter 1 Why study information economics? -- chapter 2 How to use this book -- part PART I Information as an economic good -- chapter 3 What is information? -- chapter 4 The value of information -- chapter 5 The optimal amount of information -- chapter 6 The production of information -- part PART II How the market aggregates information -- chapter 7 From information to prices -- chapter 8 A Introduction -- chapter 9 Coordination problems -- chapter 10 Learning and cascades -- chapter 11 The macroeconomics of information -- part PART III Asymmetric information -- chapter 12 The winner's curse -- chapter 13 Information and selection -- chapter 14 Optimal contracts -- chapter 15 The revelation principle -- chapter 16 A Introduction -- part PART IV The economics of self-knowledge -- chapter 17 Me, Myself, and I.