Search

Current fully-supervised facial landmark detection methods have progressed rapidly and achieved remarkable performance. However, they still suffer when coping with faces under large poses and heavy occlusions for inaccurate facial shape constraints and insufficient labeled training samples. In this paper, we propose a semi-supervised framework, i.e., a Self-Calibrated Pose Attention Network (SCPAN) to achieve more robust and precise facial landmark detection in challenging scenarios. To be specific, a Boundary-Aware Landmark Intensity (BALI) field is proposed to model more effective facial shape constraints by fusing boundary and landmark intensity field information. Moreover, a Self-Calibrated Pose Attention (SCPA) model is designed to provide a self-learned objective function that enforces intermediate supervision without label information by introducing a self-calibrated mechanism and a pose attention mask. We show that by integrating the BALI fields and SCPA model into a novel self-calibrated pose attention network, more facial prior knowledge can be learned and the detection accuracy and robustness of our method for faces with large poses and heavy occlusions have been improved. The experimental results obtained for challenging benchmark datasets demonstrate that our approach outperforms state-of-the-art methods in the literature., Accept by IEEE Transactions on Cybernetics, December 2021

Aims: This work was designed to identify the subgroup of advanced hepatocellular carcinoma (HCC) patients for whom treatments containing immune checkpoint blockers (ICBs) were most effective. Materials & methods: A meta-analysis was performed to explore the subgroup population with the greatest benefit of treatments containing ICBs. Results: A total of 2228 patients from four randomized control trials were included. Treatments containing ICBs had better overall survival, progression-free survival and higher objective response rate over treatment without ICBs. Subgroup analysis revealed that treatments containing ICBs were highly effective in improving the overall survival of males, patients with macrovascular invasion and/or extrahepatic spread and viral-related HCC patients. Conclusion: Treatments containing ICBs are more effective for males, patients with macrovascular invasion and/or extrahepatic spread and viral-related HCC patients.

Future mobile devices are anticipated to perceive, understand and react to the world on their own by running multiple correlated deep neural networks locally on-device. Yet the complexity of deep models needs to be trimmed down to fit in mobile storage and memory. Previous studies squeeze the redundancy within a single model. In this work, we reduce the redundancy across multiple models. We propose Multi-Task Zipping (MTZ), a framework to merge correlated, pre-trained deep neural networks for cross-model compression. Central in MTZ is a layer-wise neuron sharing and incoming weight updating scheme that induces a minimal change in the error function. MTZ inherits information from eachmodel and demands light retraining to re-boost the accuracy of individual tasks. MTZ supports typical network layers and applies to inference tasks with different input domains. Evaluations show that MTZ can fully merge the hidden layers of two VGG-16 network. Moreover, MTZ can effectively merge nine residual networks for diverse inference tasks and models for different input domains. Withthe joint model merged by MTZ, the latency to switch between these tasks on memory-constrained devices is reduced by 8.71.

In the process of reservoir exploitation, the range of rock stress field changes is much larger than the reservoir area where the seepage occurs; so the amount of calculation of the existing seepage stress full coupling method is often very huge. Based on the theory of seepage mechanics and rock mechanics, a coupling analysis model of reservoir multiphase seepage and stress is established, and a numerical solution is established by using finite element and finite difference methods. The evolution law of the seepage field and stress field and the change of rock mechanics parameters can be studied, with emphasis on the readjustment of rock stress distribution and its deformation characteristics under the influence of the seepage field. At the same time, to improve the calculation efficiency of numerical simulation, considering the difference between the seepage calculation area and stress calculation area, the finite element method is improved. The percolation area of the reservoir is calculated with a fine grid to obtain a more accurate distribution of underground fluid. The coarse grid is used to calculate the stress calculation area and to reduce the calculation time. The mechanical equilibrium equation in the fully coupled theory is discretized on the coarse grid by the finite element method. The mass conservation equation is discretized on the fine grid by the finite volume method. The numerical simulation of Terzaghi’s one-dimensional consolidation problem and Mandel’s two-dimensional consolidation problem shows that the calculation results of this method are in good agreement with the analytical solution. Through the numerical calculation of a two-dimensional single-phase flow single well production problem and a three-dimensional two-phase flow five-point well pattern production problem, the influence of the seepage stress coupling effect in reservoir numerical simulation is analyzed.

Objective. To investigate the clinical characteristics and pathogenic genetic mutations of a Chinese family with anterior segment mesenchymal dysgenesis and congenital posterior polar cataract. Methods. Through family investigation, the family members were examined via slit lamp anterior segment imaging and screened for eye and other diseases by eye B-ultrasound. Genetic test was performed on the blood samples of the fourth family generation (23 people) via whole exome sequencing (trio-WES) and Sanger sequencing. Results. Among the 36 members in four family generations, there were 11 living cases with different degrees of ocular abnormalities, such as cataracts, leukoplakia, and small cornea. All patients who received the genetic test had the heterozygous frameshift mutation c.640_656dup (p.G220Pfs ∗ 95) on exon 4 of the PITX3 gene. This mutation was cosegregated with the clinical phenotypes in the family and thus might be one of the genetic factors that cause the corresponding ocular abnormalities in this family. Conclusion. The congenital posterior polar cataract with or without anterior interstitial dysplasia (ASMD) of this family was inherited in an autosomal dominant manner, and the frameshift mutation (c.640_656dup) in the PITX3 gene was the cause of ocular abnormalities observed in this family. This study is of great significance for guiding prenatal diagnosis and disease treatment.

As an important part of the operating system, the file system has always received wide attention, and different types of file system architectures have been constantly proposed and practiced, this time, we mainly focus on the log structured file system (LFS) type flash friendly file system (F2FS) file system. The existing garbage collection algorithm has offered a interface for future improvement. Hence, we tried to improve existing garbage collection algorithms by implementing the working set strategy in victim select algorithm and analyze the existing deficiencies of the F2FS file system. Under the original strategy, when F2FS selects a segment, it will select the section with the fewest valid blocks. After we change it, we have completed the balance consideration of the change time, the number of valid blocks, and the proportion of changes. It is expected that the Flash device will have the longer life. By using the proposed technique, the cost of garbage collection has reduced by 10%. At the same time, based on the cold and hot data separation characteristics of the F2FS file system, we propose optimization suggestions for the cold and hot data separation strategy.

Along with the country's comprehensive strength, the people's wealth is also more and more substantial, in every aspect of life has been significantly improved, people also pay more attention to environmental protection. The harm of environmental pollution should not be underestimated. Once the environmental pollution accident occurs, it must be handled promptly, or a series of consequences are very serious. Therefore, it is necessary to study the Bayesian method of water quality emergency monitoring in the disposal of environmental pollution accidents. The purpose of this article was to solve the slow progress of water quality monitoring in the disposal of environmental pollution accidents. Through the Bayesian method of environmental pollution accident disposal in water quality emergency monitoring, the target node variables, intermediate node variables, and evidence node variables are used for analysis. Construct a Bayesian network topology. This article uses the fuzzy Bayesian network risk assessment model to provide a reference for the emergency monitoring of water quality disposal in regional environmental pollution accidents and conduct risk assessment to facilitate the next work arrangement. The results show that the Bayesian method of water quality emergency monitoring in environmental pollution accident disposal can provide effective feedback. Compared with traditional methods, the risk assessment of water quality monitoring for environmental pollution accidents is 40% faster and 10% more accurate in calculating the probability of accidents caused by basic factors. The application of the Bayesian water quality emergency monitoring method in environmental pollution accident disposal is conducive to improving the efficiency of environmental pollution disposal, in-depth analysis of the existing problems in the current environmental emergency management work, and suggestions for improving the emergency mechanism of sudden environmental pollution accident. The purpose was to provide reference for the government to properly handle sudden environmental pollution accidents.