Synativ’s Post

Synativ’s Earth Observation Week 🛰 🌍 – DAY 2 Today, in San Diego, Esri User Conference is going at full steam. It attracts many industry leaders, enthusiasts, users, and vendors all taking all things earth observation. 🏆 If you are there, here are the top 5 things we are excited about (of course it is all about getting working computer vision and deep learning models on geospatial data): - Imagery Summit https://lnkd.in/eHJgM78g - Using Large Language Models and Foundation Models by Rohit S. and Karthik Dutt https://lnkd.in/emHq4ba2 - The Day 3 Problem: Keeping Geospatial Computer Vision Models Performant by Eddy Chavarria https://lnkd.in/esvrnZuT - GeoAI Unleashed: How Geospatial is Set Free with Deep Learning by Steven P. Santovasi, GISP, CPM, CPMM https://lnkd.in/eThqQKQB - GeoAI Best Practices by by Rohit S., Karthik Dutt , and Priyanka Tuteja https://lnkd.in/e3Cn4GQx Thank you for all the sessions leaders putting in time, effort, and energy into putting these topics together. Let us know if we missed anything! #earthobservation #geospatial #computervision #deeplearning #foundationmodels #data #esri #geoai

As earth observation specialists at GeoVille Information Systems and Data Processing GmbH, our goal is to contribute to the work of our AI4Trees project partners by providing better insight into tree growth and forest changes with a perspective from outer space.   Since recent studies have shown promising results on single tree detection in dense European forests using high resolution orthophotos, we are working to integrate sub-meter resolution satellite imagery and modern machine learning techniques with the detailed ground data provided by our partners such as Umweltdata and the Bundesforschungszentrum für Wald (BFW). Our aim is to automatically detect visible tree crowns and count the number of trees in Austrian forests. Applying the method at different points in time should also reveal changes in the forest at the individual tree level, which might enable a detailed view on disturbances like clear-cuts or forest thinning. Additionally, the model will be adapted and refined for HR (Sentinel) imagery to draw conclusions on its applicability and thoroughly compare the accuracy of results considering both EO data types (VHR vs HR). At the current stage of the project, we have completed the data collection, and we are preparing the data for machine-learning. The single tree point cloud data is being transformed into planar projections of tree crown perimeters, considering obstructions by neighboring trees. This ground truth data of tree crowns will be the foundation for the next phase, where we will use it to train a U-Net based neural network for detecting trees. 🌲 #ai4trees #ai4green #sustainableforesty #climatechange

Check out this quick map I created on QGIS using Land Use Data from ESRI's Living Atlas, and a tutorial by WiseGIS Spatial Solutions Ltd. This land use/land cover (LULC) data was derived from ESA Sentinel-2 Imagery at 10m resolution. The land use types are generated by Impact Observatory's deep learning AI land classification model. More information on the data layer here: https://lnkd.in/e_E2jGym

The Geospatial Commission published today the report (there's a blog post too) of the project we've been collaborating on for several months. To add a bit of a personal view, I wrote a summary and some thoughts here also: https://lnkd.in/eGdtv87Z

What role can artificial intelligence play in long-term decision making about land use in the UK? In our latest blog we look at our partnership with The Alan Turing Institute and the prototype local land decision tool they built called DemoLand, which suggests land use scenarios to meet different policy priorities. Find out more: https://lnkd.in/eiqhV2ah

This is a fascinating read, I have been coming back to it over many days. having used remotely sensed data over many years. I was particularly interested in how medium resolution satellite data can provide value in finer context. I think it might be too aspirational, but what if there was one repository to gain insight from the hundreds of high quality, lower altitude surveys being carried out by drones everyday (on say non-secure government funded projects)?

Yesterday we published a large building footprints dataset to the Central Asia and Caucasus (CAC) GeoPortal. The dataset contains 23 million building footprints identified by Microsoft deep learning technology. Each user of the geoportal can use it for making their own insights or just downloading it. This data might be intresting to compare with cadastral data and for many other purposes. https://lnkd.in/dGQmwWXa Web map: https://lnkd.in/diq7ubeg File geodatabase: https://lnkd.in/d4NQYe3m #Centralasia #Caucasus #GIS #Geoportal #гис #geospatial #microsoft #deeplearning #remotesensing

🌐 Exciting News Alert! 🚀 Delve into the world of Advanced GIS Analysis with AI and Machine Learning! 🗺️🤖 Discover how the integration of AI and ML is reshaping Geographic Information Systems, revolutionizing spatial data analysis across various industries from urban planning to environmental management. 🌍💡 Uncover the applications of AI and ML in GIS, from land use classification to disaster management, transportation optimization, and more! 🌳🏙️🌊 Ready to explore the future of spatial analysis? Dive into our latest article now! 📈🔄 #AI #MachineLearning #GIS #SpatialAnalysis #FutureTech 🔗 Read more : https://buff.ly/3zY6R89 🌐 Stay ahead of the curve with cutting-edge technology! 🚀✨

📰 Machine Learning-Based Forest Type Mapping from Multi-Temporal Remote Sensing Data: Performance and Comparative Analysis ✍ Yusuf Ibrahim, Umar Yusuf Bagaye and Abubakar Ibrahim Muhammad This paper presents a meticulous exploration of advanced machine learning techniques for precise forest type classification using multi-temporal remote sensing data within a woodland environment. The study comprehensively evaluates a diverse range of models, ranging from advanced (ensemble) machine learning (ML) methods to several finely tuned support vector machine (SVM) variants, with a specific focus on Bayesian-optimized SVM with a radial basis function (RBF) kernel. Our findings highlight the robust performance of the Bayesian-optimized SVM, achieving a high accuracy of up to 94.27% and average precision and recall of 94.46% and 94.27%, respectively. Notably, this accuracy aligns with the levels attained by acclaimed ensemble techniques such as random forest and CatBoost while also surpassing those of XGBoost and LightGBM. These results highlight the potential of these methodologies to significantly enhance forest type mapping accuracy compared to traditional (linear) SVM and black-box neural networks. This, in turn, can enable the reliable identification and quantification of key services, including carbon storage and erosion protection, intrinsic to the forest ecosystem. The findings of the comparative study emphasize the profound impact of employing and fine-tuning ML approaches in the realm of remote sensing-based environmental analysis. 🔗 Read the paper online https://lnkd.in/g3aQNFEQ. #forestmapping #remotesensing #machinelearning #ensemblelearning #vectormachine #Bayesian #ECRS2023

Leveraging the analytical power of ArcGIS Pro alongside ArcGIS Image Analyst, Fairfax Co., VA GIS staff created a unique workflow to find hemlocks automating tree detection with a deep learning model. https://ow.ly/TPR830sJvmw