Overview
In this article, you’ll learn how to define your Area of Interest (AOI) when creating an experiment in DORA. Setting the AOI narrows the focus of your analysis by limiting the features and learning points to a specific geographic boundary.
What is an AOI?
An AOI (Area of Interest) is a defined geographic boundary where DORA will analyze input features and generate predictions. It works like a box drawn around the part of your project you want to explore.
AOIs are created from shapefiles (.shp) and must be square or rectangular in format, aligned to the cardinal directions (north-south, east-west). Each AOI must include four associated shapefile components (.shp, .shx, .dbf, .prj).
Why This Step Matters
Geoscience Perspective
Using an AOI helps geoscientists focus exploration efforts on a specific area, aligning analysis with geological boundaries, known targets, or zones of interest. It reduces the inclusion of unrelated terrain that could dilute prediction relevance.
AI Perspective
The AOI acts as a spatial filter to ensure that only relevant features and learning points are used to train and apply the model. This improves the precision, speed, and clarity of the results, and avoids introducing noisy or irrelevant data.
💡 Think Strategically About Your AOI
Before selecting or uploading an AOI, consider what you're trying to achieve:
|
Step-by-Step Instructions
Open Select Area of Interest (AOI)
From the experiment setup panel, click Step 1: Select Area of Interest (AOI).
Select an AOI File
Select the appropriate file from the dropdown.
Need to upload a new AOI? Get drag and drop instructions in this article.
Set the Resolution
Use the Height and Width sliders to adjust the resolution:
Default setting: 512px x 512px
Higher pixel counts = higher resolution (see Tips & Considerations below for more information.)
The view panel updates in real time to show the shape and grid density of your AOI
Apply AOI
Click Apply AOI to confirm your selection and proceed.
Tips & Considerations
Selecting a Quality AOI
When choosing or uploading an AOI, it's not just about size or shape — data coverage matters:
Learning Points Coverage: Make sure your learning points are well distributed across your AOI. If they cluster in just one part (like the center of a donut), the model may not train effectively.
Raster Coverage: Ensure your input rasters cover most or all of your AOI. If your AOI includes large gaps or missing values in the raster layers, DORA's model may struggle to generate reliable predictions.
Avoid Overreach: A very large AOI with poor data coverage (e.g., only 30% learning point coverage, 50% raster coverage) will likely result in weaker outcomes. Match AOI size to where your best data is.
Resolution Trade-off
Higher resolution provides more detail but increases processing time. For early-stage experiments, start with moderate resolution to speed up iterations.
Resolution | Example Size | Best For | Trade-offs |
Low | 256px x 256px | Broad, sparse, lower-quality AOIs | Fastest processing, least detail |
Medium | 512px x 512px (default) | Balanced AOIs | Good balance of speed and detail |
High | 1024px x 1024px | Small, well-mapped AOIs | Most detail, slower processing |
Common Misconception
Higher resolution does not always mean higher accuracy. If your rasters are low resolution, increasing the AOI resolution only adds interpolated pixels, not new information. In fact, this can reduce accuracy by introducing artifacts. Always choose a resolution that matches the quality of your most important input rasters.
Match Data Quality
For best results, align your AOI resolution with the resolution of your input features (see Step 2: Input Features).
Final Trade-off Reminder
Bigger AOIs capture more context, but only if your data covers that area. Smaller AOIs focus on detail, but may miss the big picture. The key is to match your AOI size to where you have strong raster and learning point coverage.
Learn More
Next Steps:
Still Have Questions?
Reach out to your dedicated DORA contact or email support@VRIFY.com for more information.