Lab 4. Preparing Training Data for Land Cover Mapping скачать в хорошем качестве

Lab 4. Preparing Training Data for Land Cover Mapping

Summary This video tutorial guides viewers through the process of creating high-quality training data in Google Colab using the Earth Engine Python API. It begins by highlighting common challenges such as class imbalance, mislabeling, spectral confusion, and the labor-intensive nature of manual data collection. The tutorial covers initializing Earth Engine, loading study area boundaries and training polygons, and checking class distribution. Sentinel-2 imagery is filtered for March–June 2024, cloud-masked, and used to create a median composite. The composite and training polygons are visualized with geemap, after which the polygons are rasterized and stratified samples are extracted. Finally, the training data is exported as a CSV and the composite as a GeoTIFF for use in classification. The video sets the foundation for the next tutorial on training and evaluating land cover models. Additional Materials: 1. Python Script https://github.com/ck1972/Geospatial-... 2. Access courses at Ai. Geelabs https://aigeolabs.com/courses/ https://aigeolabs.com/sign-up/ 3. Buy 'Explainable Machine Learning for Geospatial Data Analysis: A Data-Centric Approach' book https://aigeolabs.com/books/explainab... 4. Preparing training areas in QGIS:    • Creating Training Areas in QGIS