作者:
Chen, SiningShi, YileiXiong, ZhitongZhu, Xiao Xiang
Chair of Data Science in Earth Observation Munich80333 Germany
School of Engineering and Design Munich80333 Germany
Chair of Data Science in Earth Observation The Munich Center for Machine Learning Munich80333 Germany
Three-dimensional geoinformation is of great significance for understanding the living environment;however, 3-D perception from remote sensing data, especially on a large scale, is restricted, mainly due to the high c...
详细信息
Three-dimensional geoinformation is of great significance for understanding the living environment;however, 3-D perception from remote sensing data, especially on a large scale, is restricted, mainly due to the high costs of 3-D sensors such as light detection and ranging (LiDAR). To tackle this problem, we propose a method for monocular height estimation from optical imagery, which is currently one of the richest sources of remote sensing data. As an ill-posed problem, monocular height estimation requires well-designed networks for enhanced representations to improve the performance. Moreover, the distribution of height values is long-tailed with the low-height pixels, e.g., the background (BG), as the head, and thus, trained networks are usually biased and tend to underestimate building heights. To solve the problems, instead of formalizing the problem as a regression task, we propose HTC-DC Net following the classification-regression paradigm, with the head-tail cut (HTC) and the distribution-based constraints (DCs) as the main contributions. HTC-DC Net is composed of the backbone network as the feature extractor, the HTC-AdaBins module, and the hybrid regression process. The HTC-AdaBins module serves as the classification phase to determine bins adaptive to each input image. It is equipped with a vision transformer (ViT) encoder to incorporate local context with holistic information and involves an HTC to address the long-tailed problem in monocular height estimation for balancing the performances of foreground (FG) and BG pixels. The hybrid regression process does the regression via the smoothing of bins from the classification phase, which is trained via DCs. The proposed network is tested on three datasets of different resolutions, namely ISPRS Vaihingen (0.09 m), data Fusion Contest 19 (DFC19) (1.3 m), and Global Building Height (GBH) (3 m). The experimental results show the superiority of the proposed network over existing methods by large margins. Extensiv
This study introduces an innovative framework for Facial Movements Disorders (FMD), including eyelid movement, and mouth droopy. The system operates in real-time through front-facing laptop cameras, uploaded videos, a...
详细信息
ISBN:
(数字)9798331508944
ISBN:
(纸本)9798331508951
This study introduces an innovative framework for Facial Movements Disorders (FMD), including eyelid movement, and mouth droopy. The system operates in real-time through front-facing laptop cameras, uploaded videos, and images. The approach involves initializing the MediaPipe face landmark model, preprocessing frames/images using OpenCV, and extracting landmarks for identifying facial points. By analyzing distances, predefined thresholds, and movements are classified. We implemented a user-friendly web interface using the Streamlit library to empower users to use the implemented model easily. The system was rigorously evaluated on a labeled dataset. Evaluation metrics (accuracy, precision, recall, F1 score) were calculated. Results indicated a high accuracy of 93% for eyelid status, and 95% for mouth droopy corner.
data quality in companies is decisive and critical to the benefits their products and services can provide. However, in heterogeneous IT infrastructures where, e.g., different applications for Enterprise Resource Plan...
详细信息
Progress in the emerging fields of atomic and close-to-atomic scale manufacturing is underpinned by enhanced precision and optimization of laser-controlled nanostructuring. Understanding thin films' crystallograph...
详细信息
Breast cancer is a common cancer among women worldwide, and early diagnosis is crucial for improving the chances of successful treatment. Decision support systems (DSS) are computer-based systems that help and guide u...
详细信息
ISBN:
(数字)9798350376111
ISBN:
(纸本)9798350376128
Breast cancer is a common cancer among women worldwide, and early diagnosis is crucial for improving the chances of successful treatment. Decision support systems (DSS) are computer-based systems that help and guide users in making decisions. This paper discusses the characteristics of an optimal DSS for diagnosing breast cancer, including accuracy and reliability. This paper also introduces an optimal DSS for diagnosing breast cancer that should be able to provide an accurate, reliable, and comprehensive guide to users while being up-to-date, and flexible. The results obtained from the R-squared in the model succeeded in explaining 70% of the variation in the outputs. In contrast, the initial value of R-squared was 0.705091558 and then improved to 0.700018844. Such a system has the potential to improve the early diagnosis and treatment of breast cancer significantly and could ultimately save lives.
We begin by addressing the time-domain full-waveform inversion using the adjoint method. Next, we derive the scaled boundary semi-weak form of the scalar wave equation in heterogeneous media through the Galerkin metho...
详细信息
We begin by addressing the time-domain full-waveform inversion using the adjoint method. Next, we derive the scaled boundary semi-weak form of the scalar wave equation in heterogeneous media through the Galerkin method. Unlike conventional scaled boundary finite element formulations, the resulting system incorporates variable density and two additional terms involving its spatial derivative. As a result, the coefficient matrices are no longer constant over the elements but instead depend on the so-called radial coordinate, rendering the common solution methods such as low-frequency expansion and continued fractions—both of which assume that these matrices can be defined solely by boundary values—inapplicable. In this context, we introduce a radial discretization scheme for solving the transient scalar wave equation with spatially varying density within the framework of the scaled boundary finite element method. This approach begins with temporal discretization in the time domain, followed by spatial discretization along the radial coordinate. We employ the Newmark method for the former and a finite differencing scheme for the latter. However, the choices underlying our ansatz are made for demonstration purposes and remain flexible. Following these discretizations, we introduce an algorithmic condensation procedure to compute the dynamic stiffness matrices of the elements on the fly. Therefore, we maneuver around the need to introduce auxiliary unknowns into the global algebraic system. As a result, the optimization problem is structured in a two-level hierarchy. The higher level computes the nodal wavefields on a relatively coarse mesh, while the lower level provides the gradient of the cost function for a finely distributed set of model parameters. We obtain the Fréchet kernel by computing the zero-lag cross-correlations of the forward and adjoint wavefields, and solve the minimization problems iteratively by moving downhill on the cost function hypersurface throug
Generative Artificial Intelligence (GAI) has garnered considerable attention across disciplines such as science, digital forensics, and literature. Advanced large language modeling systems (LLMs), including ChatGPT an...
详细信息
ISBN:
(数字)9798331508944
ISBN:
(纸本)9798331508951
Generative Artificial Intelligence (GAI) has garnered considerable attention across disciplines such as science, digital forensics, and literature. Advanced large language modeling systems (LLMs), including ChatGPT and Large Language Model Meta AI (LLaMA), have become essential tools in digital forensics due to their sophisticated Natural Language Processing (NLP) capabilities. These systems enable efficient processing of extensive text datasets, sentiment analysis, and real-time threat detection. This research explores the effectiveness of AI-driven methods in digital forensics by conducting comprehensive tests on various applications, such as artifact comprehension, evidence search, and incident response. The results confirm the transformative role of ChatGPT in enhancing the speed and accuracy of investigation, as the study showed that the system can analyze data and images related to crimes and provide comprehensive reports. Not only does he process evidence, but it can also extract complete conversations related to crime, determining when it occurred and whether it was planned. The system meticulously analyzes the data sent to it to provide additional details such as possible motives and behavior of suspects. It provides investigators with an in-depth understanding that can be used at various stages of the investigation, and even in the courts as credible evidence. This approach reflects the importance of responsible adoption of the system while offering guidelines for its responsible adoption and future development.
Customer segmentation is critical for tailoring marketing campaigns to specific consumer needs, enhancing engagement, and maximizing business profitability. This study integrates Recency-Frequency (RF) analysis with t...
详细信息
ISBN:
(数字)9798331508944
ISBN:
(纸本)9798331508951
Customer segmentation is critical for tailoring marketing campaigns to specific consumer needs, enhancing engagement, and maximizing business profitability. This study integrates Recency-Frequency (RF) analysis with the K-Means clustering algorithm to classify customers of "Ramallah Outfit" into distinct behavioral segments. The Monetary (M) component was excluded because spending levels can sometimes be misleading high or low expenditures do not always reflect true loyalty or engagement. While monetary data may indicate how much a customer spends, it does not necessarily correlate with the frequency or recency of purchases, which are more reliable indicators of customer behavior and engagement. Excluding the Monetary component allows for a clearer focus on customer retention and loyalty. A high spender might make only occasional large purchases, whereas a customer with frequent smaller purchases could be more engaged and loyal. By focusing on Recency and Frequency, the analysis more accurately captures ongoing engagement, providing actionable insights for marketing. Using a unique first-hand dataset, the model achieved a clustering accuracy of 99.47%, effectively segmenting customers into actionable clusters: Gold (34.4%), Bronze (35.2%), and Silver (30.4%). The results demonstrate the effectiveness of this refined approach in optimizing marketing efforts and fostering customer loyalty.
Supplying good quality water is an important requirement for the water authority in the West Bank of Palestine. This is due to the impact of the quality of water on public health. As a result of the increasing risks o...
Supplying good quality water is an important requirement for the water authority in the West Bank of Palestine. This is due to the impact of the quality of water on public health. As a result of the increasing risks of pollution threatening vital wells field. This research is aiming to develop a decision support system (DSS) for the evaluation of water quality using the water quality index (WQI) technique. We implemented a graphical user interface that allows users to select the physiochemical and micro-biological quality parameters of the tested sample from four basic connections namely; 1) Shufat, 2) EinSamia Wells, 3) Ramallah, and 4) Hezma. Then the test results values are fed into a KNN model that predicts the water class (excellent, good, fair, poor, very poor, and unfit to be used). Based on the water class it can be determined directly if the water can be used for the human to drink or if it can be used for other purposes such as irrigation and industry.
Blood pressure problems including hypertension and hypotension are considered common among the elderly population, especially hypertension. Recently it started being common among younger adults due to many factors rel...
Blood pressure problems including hypertension and hypotension are considered common among the elderly population, especially hypertension. Recently it started being common among younger adults due to many factors related to unhealthy lifestyles, stress, or genetic factors. Besides the fact of considering hypertension as a chronic disease, it is also considered a primary or contributing cause of complicated risky health issues such as strokes, heart diseases, and chronic kidney failure. In many cases, hypertensive patients may not be aware of their problem because they don't experience any symptoms or warning signs. For this reason, it is essential to build a decision support system to identify individuals at high risk of blood pressure problems including raised blood pressure and low blood pressure. This paper proposes a decision support model for predicting blood pressure disorders by using input variables such as sex, age, body mass index (BMI), cholesterol level, heart rate, and glucose level. The decision model helps in early warning of the potential risk of hypertension or hypotension. As a result, people under potential risk are advised to measure their blood pressure regularly and take the needed precautions or medications to avoid or control this health issue. The proposed decision support model is based on using supervised machine learning classification algorithms, mainly Random Forest, Decision Tree, and XGBoost. The experimental results show that the model achieved the best performance when implemented using a Random Forest classifier with a 10-fold cross-validation method, with an accuracy of 85.81%.
暂无评论