Glossary

This glossary provides definitions for key terms and components used within the TytonAI platform to support geospatial analysis, classification, and AI training workflows.

Term	Definition
Accuracy Data	Ground-truth data used to assess the performance of classification models.
Add Band	A feature in the Upload Imagery module that allows users to include additional spectral bands (e.g. Near Infrared or Alpha) beyondthe standard RGB.
AOI (Area of Interest)	A polygon that defines the spatial extent for imagery analysis or training.
Assessment Area	Polygon(s) selected to assess classification accuracy using ground-truth points.
Auto Polygonise	The process of converting classified raster outputs into vector polygons automatically.
Band Assignment	The process of mapping specific spectral bands (e.g. Red, Green, Blue, NIR) to their correct input positions for analysis.
Band Order	The sequence in which spectral bands (e.g., Red, Green, Blue, NIR) are arranged in a raster file.
Capture Name	A label assigned to an imagery upload event or raster capture in the system.
Class Coverage	The area in square meters that each vegetation class covers within the training dataset.
Classification	The process of labeling raster imagery pixels into specific vegetation classes using AI models.
Classification Metrics	Quantitative scores (e.g., F1, precision, recall) used to evaluate model classification performance.
Clear	A button that resets a form or training configuration panel.
Clip Geometry	A shapefile (uploaded as .zip) that defines the spatial boundary to which raster imagery will be clipped prior to analysis.
Confidence Threshold	A setting during classification or training to filter predictions based on certainty level.
Deploy	The final step in submitting imagery and configurations to the system, making the data available for analysis and model training.
Deploy Imagery	The process of finalizing and activating imagery data for analysis after uploading.
Epoch	A single pass through the entire training dataset during model training.
Explorer Panel	The left-side navigation in TytonAI showing all study areas, datasets, and classification outputs.
Feature Table	A data-driven table interface showing attributes linked to spatial features.
File Browser	A user interface element that allows users to search, browse, and select previously uploaded shapefiles or imagery within the platform.
Fine-Tune	A method of training that builds on an existing pre-trained model with new localized training data.
Graph View	A visual representation of workflow steps or model configurations used as an alternative to list or panel views.
Ground Class	A default classification type representing bare ground or non-vegetated areas.
Imagery Category	Classification of imagery sources such as UAV (drone), Manned Aerial, or Satellite, used during upload metadata configuration.
Imagery Settings	Configuration options for imagery uploads, including band types and resolution.
Input Tab	A section in the workflow panel displaying all selected inputs, configurations, and metadata related to the current task.
Job History	A log of user-submitted tasks (e.g., training, classification), showing status, progress, and duration.
Layer Correction Tool	A tool used to refine or manually correct classified layers by referencing imagery or masks.
Layer Validation	A pre-deployment check ensuring that all input tiles within a layer share identical band structures and metadata.
Mask (SAM Wand)	A raster-based region generated using the Segment Anything Model to guide polygon drawing.
Model Epochs	Specific checkpoints in model training saved after a set number of iterations.
Model Library	Repository of all trained models available to the user or organization.
Monitoring Panel	A real-time dashboard that shows running or completed training and classification jobs.
Nested Study Areas	Grouped polygons under a single parent study area, allowing hierarchical organisation and batch operations.
Polygon Tool	A drawing tool used to manually outline training or testing data on imagery.
Preprocessing	The preparation phase for SAM Wand to analyze and prepare mask suggestions.
Raster Bands	The individual color or spectral layers that compose a raster image.
Raster Tile	A single .tif image chunk, typically representing part of a larger tiled raster dataset used in image uploads.
Resolution	The ground sample distance (e.g., 10cm) used to determine imagery detail.
SAM Wand	A guided AI segmentation tool that helps users extract precise vegetation polygons by placing sample points.
Scratch Train	Model training that starts from zero, without using a base model.
Selection Panel	A right-side panel that displays attributes, dimensions, and coordinate data for selected spatial features.
Select Training Areas	An option to choose regions from the current project to use for training a model.
Show Additional Options	A collapsible section in model training that reveals advanced configuration settings.
Skip Add to Project	An import setting that skips adding a classification result directly into the current project environment.
Study Area	A defined region of analysis, distinct from the training area, often used for applying classification models.
Study Area Polygon Tool	A drawing tool used to create nested polygons within a parent study area for grouping purposes.
Test Data	Ground-truth data used to evaluate the model’s classification accuracy after training.
TIF (.tif) File	A raster format for storing geospatial data like imagery or classification results.
Training Area	A polygon defining the region within which training data is created or used for model training.
Training Data	Labeled features (drawn manually or derived from tools) used to train AI classification models.
Training Datasets	An option to include data from external projects or previously saved datasets for training.
Training Preset	A saved configuration profile with common training parameters for reuse.
Training Wizard	A step-by-step interface that guides users through model configuration and launch.
UUID	A unique identifier linking geospatial features or observations, ensuring traceability across datasets.
Zoom to Layer	A map function that centers the view on the selected polygon or raster classification.