At present, text spotting in natural scenes has become one of the research hotspots. Among them, curvilinear text and long text are the main difficulties of text spotting in natural scenes. To better solve these two types of problems, we propose a novel end-to-end text spotting model. The model includes three parts: shared convolution module, text detector module and text recognizer module. For the problem of long text, we adopt the corner attention mechanism to extract the features of long text more effectively. For the problem of curve text, we feed the rectification feature map into the SA-BiLSTM decoder to recognize the curve text more effectively. More importantly, the joint optimization strategy realizes the mutual promotion function of the text detection task and the text recognition task. Experimental results on TotalText, ICDAR2015, ICDAR2013, CTW1500, COCO-Text and MLT datasets prove that our method achieves excellent performance and robustness in text spotting tasks based on end-to-end natural scenes.

No abstract

Aiming at the problem that the traditional OCR processing method ignores the inherent connection between the text detection task and the text recognition task, This paper propose a novel end-to-end text spotting framework. The framework includes three parts: shared convolutional feature network, text detector and text recognizer. By sharing convolutional feature network, the text detection network and the text recognition network can be jointly optimized at the same time. On the one hand, it can reduce the computational burden; on the other hand, it can effectively use the inherent connection between text detection and text recognition. This model add the TCM (Text Context Module) on the basis of Mask RCNN, which can effectively solve the negative sample problem in text detection tasks. This paper propose a text recognition model based on the SAM-BiLSTM (spatial attention mechanism with BiLSTM), which can more effectively extract the semantic information between characters. This model significantly surpasses state-of-the-art methods on a number of text detection and text spotting benchmarks, including ICDAR 2015, Total-Text.

No abstract

Background and purposeEarly diagnosis is important for treatment and prognosis of obstructive sleep apnea (OSA)in children. Polysomnography (PSG) is the gold standard for the diagnosis of OSA. However, due to various reasons, such as inconvenient implementation, less equipped in primary medical institutions, etc., it is less used in children, especially in young children. This study aims to establish a new diagnostic method with imaging data of upper airway and clinical signs and symptoms.MethodsIn this retrospective study, clinical and imaging data were collected from children ≤10 years old who underwent nasopharynx CT scan(low-dose protocol)from February 2019 to June 2020,including 25 children with OSA and 105 non-OSA. The information of the upper airway (A-line; N-line; nasal gap; upper airway volume; upper and lower diameter, left and right diameter and cross-sectional area of the narrowest part of the upper airway) were measured in transaxial, coronal, and sagittal images. The diagnosis of OSA and adenoid size were given according to the guidelines and consensus of imaging experts. The information of clinical signs, symptoms, and others were obtained from medical records. According to the weight of each index on OSA, the indexes with statistical significance were screened out, then were scored and summed up. ROC analysis was performed with the sum as the test variable and OSA as the status variable to evaluate the diagnostic efficacy on OSA.ResultsThe AUC of the summed scores (ANMAH score) of upper airway morphology and clinical index for the diagnosis of OSA was 0.984 (95% CI 0.964–1.000). When sum = 7 was used as the threshold (participants with sum>7 were considered to have OSA), the Youden’s index reached its maximum at which point the sensitivity was 88.0%, the specificity was 98.1%, and the accuracy was 96.2%.ConclusionThe morphological data of the upper airway based on CT volume scan images combined with clinical indices have high diagnostic value for OSA in children; CT volume scanning plays a great guiding role in the selection of treatment scheme of OSA. It is a convenient, accurate and informative diagnostic method with a great help to improving prognosis.

The fire recognition model based on deep learning can avoid many defects in the traditional method, but its construction requires a large amount of data to train the network parameters, and it takes a lot of time. In order to improve the accuracy of the model, this paper proposes a fire recognition model TNVGG-19 (Transfer learning + Newly fully connected layer module + VGG-19) with convolutional neural network based on transfer learning. First, we use the strategy of transfer learning to train the feature extraction network. Secondly, based on the VGG-19 model, this paper adds a newly designed fully connected layer module. Considering that flame data belongs to small sample data, we adopted a data augmentation strategy. Experiments show that the TNVGG-19 fire recognition model based on transfer learning proposed in this paper can effectively improve the accuracy of fire prediction and reduce the false alarm rate.

Abstract.With the increasing number of mobile applications, there has more challenging network management tasks to resolve. Users also face security issues of the mobile Internet application when enjoying the mobile network resources. Identifying applications that correspond to network traffic can help network operators effectively perform network management. The existing mobile application recognition technology presents new challenges in extensibility and applications with encryption protocols. For the existing mobile application recognition technology, there are two problems, they can not recognize the application which using the encryption protocol and their scalability is poor. In this paper, a mobile application identification method based on Hidden Markov Model(HMM) is proposed to extract the defined statistical characteristics from different network flows generated when each application starting. According to the time information of different network flows to get the corresponding time series, and then for each application to be identified separately to establish the corresponding HMM model. Then, we use 10 common applications to test the method proposed in this paper. The test results show that the mobile application recognition method proposed in this paper has a high accuracy and good generalization ability.

No abstract