Search

Many scenes need facial gender identification with good accuracy. Deep convolutional neural networks (CNN) with large set of training data normally give good accuracy, however, to achieve good accuracy with uncertain training data is a difficult task due to their lower explanation and potential information loss. Moreover, the uncertain-ties resulted from illumination, postures and facial expressions can lead to low accuracy of the classification. In this paper, a shadowed sets based two-level model for gender classification is proposed to address the problem. Deep convolutional neural networks are used as one-level classifier. Combining the concept of shadowed sets, one-level classification results are divided into three parts: accept domain, reject domain and uncertain domain. Samples in the uncertain domain are selected as uncertain facial images for two-level reclassification. Results show that the proposed method can further improve the accuracy compared with several existing state-of-the-art methods on the LFW dataset and Adience dataset.

Traditional reservoir numerical simulation software just only consider the static fractures but cannot simulate the impact of fracture expansion on well network development during long-term water injection, and the simulation results are poorly fitted to the actual production dynamic data. To address this problem, a characterization method for complex dynamic fracture networks is developed based on the embedded discrete fracture model ( ED-FM) with dynamic improvement of pre-processing algorithms. The study shows that an efficient waterflood-induced fracture simulation method is established by coupling cross-fracture discrimination criteria, which can efficiently simulate the dynamic changes of fractures during reservoir injection and development and reflect the influence of natural fractures on the main fracture propagation path; the results are consistent compared with Abaqus software, while the computational efficiency is higher. The application example shows that the method has good adaptability for numerical simulation of waterflood-induced fractures, can accurately achieve the fitting of the water cut rising rules of the field water injection well cluster, which and has good reference significance for production dynamic analysis of ultra-low permeability reservoirs and later well pattern adjustment. [ABSTRACT FROM AUTHOR]

Effective development of hot dry rock depends on complex fracture network formed by the enhanced geothermal system. Different scales of mass and heat transfer environment are formed between tight matrix and different fractures, which leads to the complexity of mass and heat transfer process of porous media. In order to analyze the influence of local thermal non-equilibrium on the mass and heat transfer process of complex fracture system, a micro pore scale mass and heat transfer model considering local non-thermal equilibrium is established. The microscopic mechanism of the effect of local non-thermal equilibrium assumption on mass and heat transfer process in matrix-fracture system is revealed. An embedded discrete fracture model considering local thermal non-equilibrium assumption and the solution method are also established. The necessity of considering local thermal non-equilibrium is analyzed by studying the sensitivity of formation physical parameters. The results show that ignoring the local thermal non-equilibrium assumption will underestimate the rock temperature near the fracture zone and overestimate the fluid temperature in the heat transfer front zone. The effect of local thermal non-equilibrium on the enhanced geothermal system is mainly reflected in the early stage of injection. The higher the injection intensity, the lower the matrix fracture permeability, the greater the matrix fracture porosity difference, the greater the rock thermal diffusion coefficient, the smaller the convective heat transfer coefficient, the more it is necessary to consider the local thermal non-equilibrium assumption. The research results can provide reference for the collection and collation of field data. [ABSTRACT FROM AUTHOR]