基于改进DenseNet网络的书法字体识别算法

汉字书法是中华传统文化的代表,但是,由于书法字体具有风格迥异、结构复杂、变形繁多等特点,给大众学习和欣赏书法带来了极大障碍.为了解决普通老百姓解读书法作品的困难,提出一种基于改进DenseNet网络的书法字体识别算法,设计区域权值比例池化规则替换传统DenseNet网络的最大池化和平均池化规则,采用Nadam算法优化模...     

样本仿真结合迁移学习的声呐图像水雷检测

水雷检测对于国防安全具有重要意义,然而,由于水下目标声呐成像实验代价较大,通常难以获得足够的水雷声呐图像样本,因此导致难以应用深度神经网络提高水雷等目标的检测精度。针对这一问题,提出样本仿真结合迁移学习的侧扫声呐图像水雷目标检测与识别方法。首先,根据侧扫声呐成像机理,建立水雷目标的仿真模型,进而仿真得到大量水雷目标样本;然后,采用大型广源域数据集ImageNet对深度卷积神经网络进行预训练,再用真实水雷样本和仿真水雷样本对深度卷积神经网络进行微调以适应水雷目标;最后,将微调后的深度卷积神经网络作为目标检测的基准网络,并进行目标检测训练;采用真实的水下水雷声呐图像数据对训练完成的网络进行验证和比较。实验结果表明,提出的基于样本仿真和迁移学习的侧扫声呐图像水雷目标检测方法能够更好地检测水雷目标,优于传统的特征提取及检测方法及只采用真实样本进行训练的检测方法,对于水下目标检测具有借鉴意义。     

Modeling and recognition of cursive words with hidden Markov models

In this paper, a new method for modeling and recognizing cursive words with hidden Markov models (HMM) is presented. In the proposed method, a sequence of thin fixed-width vertical frames are extracted from the image, capturing the local features of the handwriting. By quantizing the feature vectors of each frame, the input word image is represented as a Markov chain of discrete symbols. A handwritten word is regarded as a sequence of characters and optional ligatures. Hence, the ligatures are also explicitly modeled. With this view, an interconnection network of character and ligature HMMs is constructed to model words of indefinite length. This model can ideally describe any form of handwritten words, including discretely spaced words, pure cursive words and unconstrained words of mixed styles. Experiments have been conducted with a standard database to evaluate the performance of the overall scheme. The performance of various search strategies based on the forward and backward score has been compared. Experiments on the use of a preclassifier based on global features show that this approach may be useful for even large-vocabulary recognition tasks.     

Rotation,Translation and Scale Invariant Sign Word Recognition Using Deep Learning

Communication between people with disabilities and people who do not understand sign language is a growing social need and can be a tedious task. One of the main functions of sign language is to communicate with each other through hand gestures. Recognition of hand gestures has become an important challenge for the recognition of sign language. There are many existing models that can produce a good accuracy, but if the model test with rotated or translated images, they may face some difficulties to make good performance accuracy. To resolve these challenges of hand gesture recognition, we proposed a Rotation, Translation and Scale-invariant sign word recognition system using a convolutional neural network (CNN). We have followed three steps in our work: rotated, translated and scaled (RTS) version dataset generation, gesture segmentation, and sign word classification. Firstly, we have enlarged a benchmark dataset of 20 sign words by making different amounts of Rotation, Translation and Scale of the original images to create the RTS version dataset. Then we have applied the gesture segmentation technique. The segmentation consists of three levels, i) Otsu Thresholding with YCbCr, ii) Morphological analysis: dilation through opening morphology and iii) Watershed algorithm. Finally, our designed CNN model has been trained to classify the hand gesture as well as the sign word. Our model has been evaluated using the twenty sign word dataset, five sign word dataset and the RTS version of these datasets. We achieved 99.30% accuracy from the twenty sign word dataset evaluation, 99.10% accuracy from the RTS version of the twenty sign word evolution, 100% accuracy from the five sign word dataset evaluation, and 98.00% accuracy from the RTS version five sign word dataset evolution. Furthermore, the influence of our model exists in competitive results with state-of-the-art methods in sign word recognition.     

A recursive framework for expression recognition: from web images to deep models to game dataset

In this paper, we propose a recursive framework to recognize facial expressions from images in real scenes. Unlike traditional approaches that typically focus on developing and refining algorithms for improving recognition performance on an existing dataset, we integrate three important components in a recursive manner: facial dataset generation, facial expression recognition model building, and interactive interfaces for testing and new data collection. To start with, we first create candid images for facial expression (CIFE) dataset. We then apply a convolutional neural network (CNN) to CIFE and build a CNN model for web image expression classification. In order to increase the expression recognition accuracy, we also fine-tune the CNN model and thus obtain a better CNN facial expression recognition model. Based on the fine-tuned CNN model, we design a facial expression game engine and collect a new and more balanced dataset, GaMo. The images of this dataset are collected from the different expressions our game users make when playing the game. Finally, we run yet another recursive step—a self-evaluation of the quality of the data labeling and propose a self-cleansing mechanism for improve the quality of the data. We evaluate the GaMo and CIFE datasets and show that our recursive framework can help build a better facial expression model for dealing with real scene facial expression tasks.     

基于图像聚类的交通标志CNN快速识别算法

为了提高交通标志图像识别的准确性和实时性,提出一种基于图像聚类的交通标志CNN快速识别算法。利用图像聚类算法对原始数据集进行样本优化;采用多种图像预处理操作使样本整体质量进一步提升;构造了深度为9的CNN结构,通过多次训练得到最终的网络模型,将待识别的图像输入到CNN模型来实现自动识别。在德国交通标志数据集（German traffic sign recognition benchmark, GTSRB）和比利时交通标志数据集（Belgium traffic sign dataset, BTSD）上证明了算法的有效性,单张图片的识别速度只需0.2 s,识别精度高达98.5%以上。本算法具有识别速度快、准确率高的特点,可为智能驾驶的可靠性和安全性提供理论依据和技术支持。     

基于CNN-BLSTM-CRF模型的生物医学命名实体识别

命名实体识别是自然语言处理任务的重要步骤。近年来,不依赖人工特征的神经网络在新闻等通用领域命名实体识别方面表现出了很好的性能。然而在生物医学领域,许多实验表明基于领域知识的人工特征对于神经网络模型的结果影响很大。因此,如何在不依赖人工特征的情况下获得较好的生物医学命名实体识别性能是有待解决的问题。该文提出一种基于CNN-BLSTM-CRF的神经网络模型。首先利用卷积神经网络(CNN)训练出单词的具有形态特征的字符级向量,并从大规模背景语料训练中得到具有语义特征信息的词向量,然后将二者进行组合作为输入,再构建适合生物医学命名实体识别的BLSTM-CRF深层神经网络模型。实验结果表明,不依赖任何人工特征,该文方法在Biocreative Ⅱ GM和JNLPBA2004生物医学语料上都达到了目前最好的结果,F-值分别为89.09%和74.40%。     

A novel attention fusion network-based framework to ensemble the predictions of CNNs for lymph node metastasis detection

Diagnosis of lymph node metastases is a challenging task for pathologists, involving an extensive screening of the pathological scans. Automating diagnostic processes reduces the workload of pathologists and yields high accuracy by the virtue of advances in technology. In this study, a novel ensemble-based framework is proposed for the classification of lymph node metastases. The proposed ensemble framework comprises different pre-trained CNN models such as DenseNet201, InceptionV3 and ResNeXt-50. In the proposed framework, an attention fusion network is utilized to amalgamate the predictions of the individual models. The proposed framework achieves an AUC-ROC of 0.9816 which surpasses the highest AUC-ROC achieved by the conventional approaches on the PCam benchmark dataset.

     

FCN and LSTM Based Computer Vision System for Recognition of Vehicle Type,License Plate Number,and Registration Country

We propose an advanced Automatic number-plate recognition (ANPR) system, which not only recognizes the number and the issuing state, but also the type and location of the vehicle in the input image. The system is based on a combination of existing methods, modifications to neural network architectures and improvements in the training process. The proposed system uses machine-learning approach and consists of three main parts: segmentation of input image by Fully Convolutional Network for localization of license plate and determination of vehicle type; recognition of the characters of the localized plate by a Maxout CNN and LSTM; determination of the state that has issued the license plate by a CNN. The training of these neural network models is accomplished using a manually labeled custom dataset, which is expanded with data augmented techniques. The resulting system is capable of localizing and classifying multiple types of vehicles (including motorcycles and emergency vehicles) as well as their license plates. The achieved precision of the localization is 99.5%. The whole number recognition accuracy is 96.7% and character level recognition accuracy is 98.8%. The determination of issuing state is precise in 92.8% cases.     

Evaluation of Deep Learning Models for Person Authentication Based on Touch Gesture

Touch gesture biometrics authentication system is the study of user's touching behavior on his touch device to identify him. The features traditionally used in touch gesture authentication systems are extracted using hand-crafted feature extraction approach. In this work, we investigate the ability of Deep Learning (DL) to automatically discover useful features of touch gesture and use them to authenticate the user. Four different models are investigated Long-Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Network (CNN) combined with LSTM (CNN-LSTM), and CNN combined with GRU(CNN-GRU). In addition, different regularization techniques are investigated such as Activity Regularizer, Batch Normalization (BN), Dropout, and LeakyReLU. These deep networks were trained from scratch and tested using TouchAlytics and BioIdent datasets for dynamic touch authentication. The result reported in terms of authentication accuracy, False Acceptance Rate (FAR), False Rejection Rate (FRR). The best result we have been obtained was 96.73%, 96.07% and 96.08% for training, validation and testing accuracy respectively with dynamic touch authentication system on TouchAlytics dataset with CNN-GRU DL model, while the best result of FAR and FRR obtained on TouchAlytics dataset was with CNN-LSTM were FAR was 0.0009 and FRR was 0.0530. For BioIdent dataset the best results have been obtained was 84.87%, 78.28% and 78.35% for Training, validation and testing accuracy respectively with CNN-LSTM model. The use of a learning based approach in touch authentication system has shown good results comparing with other state-of-the-art using TouchAlytics dataset.     

A SVM-based cursive character recognizer

This paper presents a cursive character recognizer, a crucial module in any cursive word recognition system based on a segmentation and recognition approach. The character classification is achieved by using support vector machines(SVMs) and a neural gas. The neural gas is used to verify whether lower and upper case version of a certain letter can be joined in a single class or not. Once this is done for every letter, the character recognition is performed by SVMs. A database of 57 293 characters was used to train and test the cursive character recognizer. SVMs compare notably better, in terms of recognition rates, with popular neural classifiers, such as learning vector quantization and multi-layer-perceptron. SVM recognition rate is among the highest presented in the literature for cursive character recognition.     

并行生成网络的红外—可见光图像转换

目的 针对现有图像转换方法的深度学习模型中生成式网络（generator network）结构单一化问题，改进了条件生成式对抗网络（conditional generative adversarial network，CGAN）的结构，提出了一种融合残差网络（ResNet）和稠密网络（DenseNet）两种不同结构的并行生成器网络模型。方法 构建残差、稠密生成器分支网络模型，输入红外图像，分别经过残差、稠密生成器分支网络各自生成可见光转换图像，并提出一种基于图像分割的线性插值算法，将各生成器分支网络的转换图像进行融合，获取最终的可见光转换图像；为防止小样本条件下的训练过程中出现过拟合，在判别器网络结构中插入dropout层；设计最优阈值分割目标函数，在并行生成器网络训练过程中获取最优融合参数。结果 在公共红外-可见光数据集上测试，相较于现有图像转换深度学习模型Pix2Pix和CycleGAN等，本文方法在性能指标均方误差（mean square error，MSE）和结构相似性（structural similarity index，SSIM）上均取得显著提高。结论 并行生成器网络模型有效融合了各分支网络结构的优点，图像转换结果更加准确真实。