RESUMO
Wall segmentation is a special case of semantic segmentation, and the task is to classify each pixel into one of two classes: wall and no-wall. The segmentation model returns a mask showing where objects like windows and furniture are located, as well as walls. This article proposes the module's structure for semantic segmentation of walls in 2D images, which can effectively address the problem of wall segmentation. The proposed model achieved higher accuracy and faster execution than other solutions. An encoder-decoder architecture of the segmentation module was used. Dilated ResNet50/101 network was used as an encoder, representing ResNet50/101 network in which dilated convolutional layers replaced the last convolutional layers. The ADE20K dataset subset containing only interior images, was used for model training, while only its subset was used for model evaluation. Three different approaches to model training were analyzed in the research. On the validation dataset, the best approach based on the proposed structure with the ResNet101 network resulted in an average accuracy at the pixel level of 92.13% and an intersection over union (IoU) of 72.58%. Moreover, all proposed approaches can be applied to recognize other objects in the image to solve specific tasks.
RESUMO
In this paper, the influence of the variation in transformer oil temperature on the accuracy of the all-acoustic non-iterative method for partial discharge location in a power transformer is researched. The research can improve power transformers' testing and monitoring, particularly given the large transformer oil temperature variations during real-time monitoring. The research is based on quantifying the contribution of oil temperature to the standard combined measurement uncertainty of the non-iterative algorithm by using analytical, statistical, and Monte Carlo methods. The contribution can be quantified and controlled. The contribution varied significantly with different mutual placements of partial discharge and acoustic sensors. The correlation between the contribution and the mean distance between partial discharge and acoustic sensors was observed. Based on these findings, the procedure to quantify and control the contribution in practice was proposed. The procedure considers the specificity of the method's mathematical model (the assumption that the oil temperature is constant), the non-iterative algorithm's nonlinearity, and the large variations in transformer oil temperature. Existing studies did not consider the significant effect of the oil temperature on the combined measurement uncertainty of partial discharge location influenced by those phenomena. The research is limited to partial discharge located in the transformer oil.
RESUMO
Choosing a comprehensive and cost-effective way of articulating and annotating the sentiment of a text is not a trivial task, particularly when dealing with short texts, in which sentiment can be expressed through a wide variety of linguistic and rhetorical phenomena. This problem is especially conspicuous in resource-limited settings and languages, where design options are restricted either in terms of manpower and financial means required to produce appropriate sentiment analysis resources, or in terms of available language tools, or both. In this paper, we present a versatile approach to addressing this issue, based on multiple interpretations of sentiment labels that encode information regarding the polarity, subjectivity, and ambiguity of a text, as well as the presence of sarcasm or a mixture of sentiments. We demonstrate its use on Serbian, a resource-limited language, via the creation of a main sentiment analysis dataset focused on movie comments, and two smaller datasets belonging to the movie and book domains. In addition to measuring the quality of the annotation process, we propose a novel metric to validate its cost-effectiveness. Finally, the practicality of our approach is further validated by training, evaluating, and determining the optimal configurations of several different kinds of machine-learning models on a range of sentiment classification tasks using the produced dataset.
Assuntos
Linguística/métodos , Algoritmos , Humanos , Aprendizado de Máquina , Processamento de Linguagem Natural , Sérvia , Mídias SociaisRESUMO
AIMS AND OBJECTIVES: To develop and validate a reliable instrument that can measure fear of hospitalisation experienced by outpatients. BACKGROUND: After having a diagnosis established, some patients experience sense of fear, unpleasantness and embarrassment due to the possibility to be admitted to a hospital. Currently, there is no available instrument for measuring fear of hospitalisation. DESIGN: Cross-sectional study for assessing reliability and validity of a questionnaire. METHOD: The questionnaire with 17 items and answers according to the Likert scale was developed during two brainstorming sessions of the research team. Its reliability, validity and temporal stability were tested on the sample of 330 outpatients. The study was multicentric, involving patients from seven cities and three countries. RESULTS: Fear of hospitalisation scale showed satisfactory reliability, when rated both by the investigators (Cronbach's alpha .799) and by the patients themselves (Cronbach's alpha .760). It is temporally stable, and both divergent and convergent validity tests had good results. Factorial analysis revealed three domains: fear of being injured, trust to medical staff and fear of losing privacy or autonomy. CONCLUSIONS: This study developed new reliable and valid instrument for measuring fear of hospitalisation. RELEVANCE TO CLINICAL PRACTICE: Identification of patients with high level of fear of hospitalisation by this instrument should help clinicians to administer measures which may decrease fear and prevent avoidance of healthcare utilisation.
Assuntos
Medo/psicologia , Hospitalização , Pacientes Ambulatoriais/psicologia , Inquéritos e Questionários/normas , Adulto , Estudos Transversais , Feminino , Humanos , Masculino , Pessoa de Meia-Idade , Psicometria , Reprodutibilidade dos TestesRESUMO
BACKGROUND/AIMS: Predicting technical difficulties in laparoscopic cholecystectomy (LC) in a small regional hospital increases efficacy, cost-benefit and safety of the procedure. The aim of the study was to assess whether it is possible to accurately predict a difficult LC (DLC) in a small regional hospital based only on the routine available clinical work-up parameters (patient history, ultrasound examination and blood chemistry) and their combinations. METHODOLOGY: A prospective, cohort, of 369 consecutive patients operated by the same surgeon was analyzed. Conversion rate was 10 (2.7%). DLC was registered in 55 (14.90%). Various data mining techniques were applied and assessed. RESULTS: Seven significant predictors of DLC were identified: i) shrunken (fibrotic) gallbladder (GB); ii) ultrasound (US) GB wall thickness >4 mm; iii) >5 attacks of pain lasting >5 hours; iv) WBC >10x109 g/L; v) pericholecystic fluid; vi) urine amylase >380 IU/L, and vii) BMI >30kg/m2. Bayesian network was selected as the best classifier with accuracy of 94.57, specificity 0.98, sensitivity 0.77, AUC 0.96 and F-measure 0.81. CONCLUSION: It is possible to predict a DLC with high accuracy using data mining techniques, based on routine preoperative clinical parameters and their combinations. Use of sophisticated diagnostic equipment is not necessary.