دانلود کتاب کاوش در GeoAI

Geospatial artificial intelligence (GeoAI) is the fusion of arti􀄩cial intelligence (AI) with geospatial data, science, and technology. In ArcGIS , the GeoAI capability includes tools and models that help users automate geospatial data extraction and enhance their spatial analyses. With the growing availability of remotely sensed imagery, sensors, and other geospatial datasets, GeoAI is becoming increasingly important to organizations looking to improve understanding, find deeper insights in their data, and solve complex geospatial problems. At its base, GeoAI includes the application of deep learning techniques to automate geospatial data extraction from both structured and unstructured data. For example, deep learning can be used in ArcGIS to extract geospatial information from imagery, 3D, and lidar point cloud data using object detection, pixel classification, object tracking, and change detection techniques. Deep learning and large language model (LLM) integration can also be used to extract geospatial data from unstructured text using a variety of natural language processing techniques. GeoAI also encompasses machine learning and deep learning techniques for solving spatial problems through the analysis of vector, tabular, and time series data. Examples include pattern detection and clustering, predictive analysis, and spatiotemporal forecasting. GeoAI tools and techniques are embedded throughout a variety of applications and experiences in ArcGIS. You can train your own custom machine learning and deep learning models using geoprocessing tools and ® wizards in ArcGIS Pro or train them programmatically using the arcgis.learn module of the ArcGIS API for Python. ArcGIS Enterprise users can train their own models using Deep Learning Studio, a web app that provides a multiuser, project-based environment for teams to collaborate on GeoAI projects in a web browser with no software installation required. In addition to being able to train your own GeoAI models, you can also use pretrained GeoAI models available in ArcGIS Living Atlas of the World for a wide variety of specific tasks over different geographies. These pretrained models help reduce some of the time and manual effort in collecting training data, as well as the amount of compute resources required to train your own models. They can be used out of the box in ArcGIS Pro, Deep Learning Studio, and Map Viewer, and many of them can be 􀄩ne-tuned with your training data to specific geographies. Currently, more than 100 pretrained models are available for tasks, including extraction of roads, trees, building footprints, cars, water bodies, solar panels, and swimming pools from imagery, as well as land-cover classification and lidar point cloud classification of buildings, power lines, and trees. Other capabilities include object tracking in video, optical character recognition (OCR), time series forecasting, and many more such tasks. The ArcGIS pretrained models also include vision language models that integrate third-party AI services from OpenAI and Meta, as well as a seamless integration with Hugging Face Hub for a variety of computer vision and text analysis tasks. In the following chapters, you’ll get hands-on experience with using and finetuning ArcGIS pretrained models and training your own GeoAI models for imagery and lidar workflows and for performing predictive spatial analysis. With this guidance, you can explore for yourself the power of GeoAI to yield deeper insights from geospatial data.