基于改进MobileFaceNet的人脸识别方法北大核心

针对目前大多数人脸识别算法参数多、计算量大,难以部署到移动端和嵌入式设备中的问题,提出了一种基于改进MobileFaceNet的人脸识别方法。通过对MobileFaceNet模型结构的调整,将bottleneck模块优化为sandglass模块,改良深度卷积和逐点卷积的相对位置,适当增大sandglass模块的输出通道数,从而减少特征压缩时的信息丢失,增强人脸空间特征的提取。实验结果表明:改进后的方法在LFW测试数据集上准确率达99.15%,模型大小和计算量分别仅为原算法的61%和45%,验证了所提方法的有效性。     

基于卷积神经网络的5G蜂窝网络无线定位方法

5G蜂窝网络发展迅猛，其覆盖面积将逐渐增大，因此使用5G蜂窝网络进行定位是有研究潜力的研究方向。本文提出一种新的深度学习技术来实现高效、高精度和低占用的定位，以代替传统指纹定位过程中繁重的指纹库生成以及距离计算。该方法建立了一个特殊的卷积神经网络，并根据5G天线信号的接收信号强度指示、相位和到达角等特征量，选择合适的输入数据格式构造样本组建训练集，对该卷积神经网络进行训练。训练得到的卷积神经网络可以替代指纹定位中的庞大指纹库，非常有利于直接在5G移动设备端实现定位。虽然卷积神经网络在训练过程中需要大量时间，但在训练完毕后直接进行分类定位的速度非常快，可以保障定位实现的实时性。本文所实现的卷积神经网络权重与偏置所占内存不到0.5 MB，且能够在实际应用环境中以95%的定位准确率以及0.1 m的平均定位精度实现高精度定位。     

Neural Style Transfer for image within images and conditional GANs for destylization

In this paper, the feature representation of an image by CNN is used to hide the secret image into the cover image. The style of the cover image hides the content of the secret image and produce a stego image using Neural Style Transfer (NST) algorithm, which resembles the cover image and also contains the semantic content of secret image. The main technical contributions are to hide the content of the secret image in the in-between hidden layered style features of the cover image, which is the first of its kind in the present state-of-art-technique. Also, to recover the secret image from the stego image, destylization is done with the help of conditional generative adversarial networks (GANs) using Residual in Residual Dense Blocks (RRDBs). Further, stego images from different layer combinations of content and style features are obtained and evaluated. Evaluation is based on the visual similarity and quality loss between the cover-stego pair and the secret-reconstructed secret pair of images. From the experiments, it has been observed that the proposed algorithm has 43.95 dB Peak Signal-to-Noise Ratio (PSNR)), .995 Structural Similarity Index (SSIM), and .993 Visual Information Fidelity (VIF) for the ImageNet dataset. The proposed algorithm is found to be more robust against StegExpose than the traditional methods.     

基于人工神经网络的智能化架空线路应变快速解调方法北大核心

为了提高智能化光纤复合架空线路态势感知的实时性,将人工神经网络方法应用于光纤沿线应变解调,确定了神经网络的结构。编程实现了基于洛伦兹模型的最小二乘谱拟合方法和神经网络方法,采用不同信噪比和布里渊频移的布里渊谱训练神经网络,将它们应用于某光纤复合架空线路沿线光纤应变的测量,从不同角度比较了两种方法的计算结果。计算结果表明,神经网络方法能有效获得光纤沿线的布里渊频移进而获得应变,具有与谱拟合方法相似的准确性,但应变解调时间仅约为谱拟合方法的1/20000。研究结果为提高智能光纤复合架空线路态势感知的实时性提供了参考。     

Application of artificial neural network method to predict the breakage properties of PGE bearing chromite ore

In the present investigation, systematic grinding experiments were conducted in a laboratory ball mill to determine the breakage properties of low-grade PGE bearing chromite ore. The population balance modeling technique was used to study the breakage parameters such as primary breakage distribution (B_{i, j}) and the specific rates of breakage (S_i). The breakage and selection function values were determined for six feed sizes. The results stated that the breakage follows the first-order grinding kinetics for all the feed sizes. It was observed that the coarser feed sizes exhibit higher selection function values than the finer feed size. Further, an artificial neural network was used to predict breakage characteristics of low-grade PGE bearing chromite ore. The predicted results obtained from the neural network modeling were close to the experimental results with a correlation of determination R² = 0.99 for both product size and selection function.     

Learn decision trees with deep visual primitives

In this paper, we strive to propose a self-interpretable framework, termed PrimitiveTree, that incorporates deep visual primitives condensed from deep features with a conventional decision tree, bridging the gap between deep features extracted from deep neural networks (DNNs) and trees’ transparent decision-making processes. Specifically, we utilize a codebook, which embeds the continuous deep features into a finite discrete space (deep visual primitives) to distill the most common semantic information. The decision tree adopts the spatial location information and the mapped primitives to present the decision-making process of the deep features in a tree hierarchy. Moreover, the trained interpretable PrimitiveTree can inversely explain the constituents of the deep features, highlighting the most critical and semantic-rich image patches attributing to the final predictions of the given DNN. Extensive experiments and visualization results validate the effectiveness and interpretability of our method.     

基于音频特征的工程车辆工况识别研究

高效率地使用工程车辆是工程项目管理中节约成本的有效方法，无人监管环境下工程车辆的工况识别，是实现工程车辆高效率使用的有效手段。目前以GPS等技术为核心的车辆智能管理系统未对工程车辆进行工况识别，提出一种基于GRU循环神经网络的工程车辆工况识别方法，通过对工程车辆在不同工况下产生的音频信号进行分析，从中提取Mel倒谱系数作为主要特征，构建GRU循环神经网络模型进行训练和识别。实验结果表明，该方法可以实现对工程车辆工况的有效识别。     

A novel adaptive sampling based methodology for feasible region identification of compute intensive models using artificial neural network

Identification of feasible region of operations in multivariate processes is a problem of interest in several fields. This is particularly challenging when the process model is black-box in nature and/or is computationally expensive, as analytical solutions are not available and the number of possible model evaluations is limited. An efficient methodology is required to identify samples where the model is evaluated for developing a computationally efficient surrogate model. In this work, an artificial neural network based surrogate model is proposed which is integrated with a statistical-based approach (Jack-knifing) to estimate the variance of the surrogate model prediction. This allows implementation of an adaptive sampling approach where new samples are identified close to the feasible region boundary or in regions of high prediction uncertainty. The proposed approach performs better than a previously published kriging based method for different dimensionality case studies.     

Enhanced wide-activated residual network for efficient and accurate image deblocking

To save bandwidth and storage space as well as speed up data transmission, people usually perform lossy compression on images. Although the JPEG standard is a simple and effective compression method, it usually introduces various visually unpleasing artifacts, especially the notorious blocking artifacts. In recent years, deep convolutional neural networks (CNNs) have seen remarkable development in compression artifacts reduction. Despite the excellent performance, most deep CNNs suffer from heavy computation due to very deep and wide architectures. In this paper, we propose an enhanced wide-activated residual network (EWARN) for efficient and accurate image deblocking. Specifically, we propose an enhanced wide-activated residual block (EWARB) as basic construction module. Our EWARB gives rise to larger activation width, better use of interdependencies among channels, and more informative and discriminative non-linearity activation features without more parameters than residual block (RB) and wide-activated residual block (WARB). Furthermore, we introduce an overlapping patches extraction and combination (OPEC) strategy into our network in a full convolution way, leading to large receptive field, enforced compatibility among adjacent blocks, and efficient deblocking. Extensive experiments demonstrate that our EWARN outperforms several state-of-the-art methods quantitatively and qualitatively with relatively small model size and less running time, achieving a good trade-off between performance and complexity.     

Predicting the mechanical properties of Cu–Al2O3 nanocomposites using machine learning and finite element simulation of indentation experiments

Micromechanics model, finite element (FE) simulation of microindentation and machine learning were deployed to predict the mechanical properties of Cu–Al₂O₃ nanocomposites. The micromechanical model was developed based on the rule of mixture and grain and grain boundary sizes evolution to predict the elastic modulus of the produced nanocomposites. Then, a FE model was developed to simulate the microindentation test. The input for the FE model was the elastic modulus that was computed using the micromechanics model and wide range of yield and tangent stresses values. Finally, the output load-displacement response from the FE model, the elastic modulus, the yield and tangent strengths used for the FE simulations, and the residual indentation depth were used to train the machine learning model (Random vector functional link network) for the prediction of the yield and tangent stresses of the produced nanocomposites. Cu–Al₂O₃ nanocomposites with different Al₂O₃ concentration were manufactured using insitu chemical method to validate the proposed model. After training the model, the microindentation experimental load-displacement curve for Cu–Al₂O₃ nanocomposites was fed to the machine learning model and the mechanical properties were obtained. The obtained mechanical properties were in very good agreement with the experimental ones achieving 0.99 coefficient of determination R2 for the yield strength.