Semantic Segmentation in QGIS

Introduction¶

Learning Objectives¶

Creating Training Data¶

Sample Data¶

Input Raster and Vector Labels¶

Tile Size and Stride¶

Output Directory and Export¶

Training a Model¶

Images and Labels Directories¶

Model Architectures¶

Encoder Backbones¶

Training Hyperparameters¶

Starting Training¶

Running Inference¶

Input Configuration¶

Inference Parameters¶

Vectorization and Post-Processing¶

Troubleshooting Common Problems¶

Key Takeaways¶

Exercises¶

Exercise 1: End-to-End Building Segmentation¶

Exercise 2: Architecture Comparison¶

Exercise 3: Multi-Class Segmentation¶

Exercise 4: Geographic Transfer¶