Designing and optimizing complex scientific code for new computing architectures is a challenging task. To address this issue in the E3SM land model (ELM) development, we developed a software tool called SPEL, which f...Designing and optimizing complex scientific code for new computing architectures is a challenging task. To address this issue in the E3SM land model (ELM) development, we developed a software tool called SPEL, which facilitates code generation, verification, and performance tuning using compiler directives within a Function Unit Test framework. In this paper, we present a SPEL extension that leverages the version control system (e.g., Git) to autonomous code generation and demonstrate its application to continuous code integration and development of the ELM software system. The study can benefit the scientific software development community.展开更多
This report presented a method that uses deep computing and stochastic gradient descent algorithm to automatically detect building from satellite images. In this method, a convolutional neural network architecture cal...This report presented a method that uses deep computing and stochastic gradient descent algorithm to automatically detect building from satellite images. In this method, a convolutional neural network architecture called U-Net was trained to highlight the building pixels from the rest of the image. This method applied a binary cross-entropy loss function, used ADAM algorithm for gradient descent optimization, and adopted interaction-over-union for accuracy measurement. Continuous loss decreases and accuracy increases were observed during the training and validation. Finally, the visualization of the predicted masks from the trained model after 20 epochs proved that the U-Net model delivers over 60% Intersection over Union accuracy results for detecting buildings from satellite images.展开更多
文摘Designing and optimizing complex scientific code for new computing architectures is a challenging task. To address this issue in the E3SM land model (ELM) development, we developed a software tool called SPEL, which facilitates code generation, verification, and performance tuning using compiler directives within a Function Unit Test framework. In this paper, we present a SPEL extension that leverages the version control system (e.g., Git) to autonomous code generation and demonstrate its application to continuous code integration and development of the ELM software system. The study can benefit the scientific software development community.
文摘This report presented a method that uses deep computing and stochastic gradient descent algorithm to automatically detect building from satellite images. In this method, a convolutional neural network architecture called U-Net was trained to highlight the building pixels from the rest of the image. This method applied a binary cross-entropy loss function, used ADAM algorithm for gradient descent optimization, and adopted interaction-over-union for accuracy measurement. Continuous loss decreases and accuracy increases were observed during the training and validation. Finally, the visualization of the predicted masks from the trained model after 20 epochs proved that the U-Net model delivers over 60% Intersection over Union accuracy results for detecting buildings from satellite images.
