Skip to main content

Training Data Requirements & Considerations

Fine-tuning requires labeled data that links a location, a time, and a target variable. These labels are used alongside model input data (e.g., Sentinel-2 imagery) to teach the model what patterns correspond to the phenomenon you want to detect or predict. This data may come from open sources or from your own organization.

Required fieldDescriptionExample
LocationWhere the label was observed, as a point or polygon in WGS84 (EPSG:4326) coordinatesA field observation at latitude 45.52, longitude -122.68
TimeWhen the label applies, as a single observation date or a date range2023-07-14, or 2023-06-01 to 2023-08-31
Target variableWhat you want the model to detect or predict, as a category, a number, or a booleanBurned, 0.32 (soil moisture), or true

Requirements at a glance​

  • Format: CSV (points only) or GeoJSON (points or polygons, including multipolygons). GeoJSON files can contain a FeatureCollection or individual Features.
  • Coordinate system: WGS84 (EPSG:4326), with coordinates in decimal-degree latitude and longitude. In CSV files, latitude and longitude must be in separate columns. Studio does not reproject vector data, so data in another coordinate system will be placed incorrectly.
  • Date: Each point or polygon must have either a single observation date (e.g., the date of a field observation) or a date range the label applies to (e.g., an aggregated measurement period). A year (e.g., 2023) covers January 1 to December 31, and a month (e.g., 2023-07, 07/2023, or July 2023) covers its first to last day.
  • Label: Each point or polygon must have a label value:
    • Category: a set of nouns (e.g., buildings, water, trees)
    • Number: measurement values (e.g., moisture content)
    • True or false: boolean statements

Once your dataset meets these requirements, you will be able to upload it to OlmoEarth Studio.

Training Data Considerations​

"How much data do I need?" is a common question. The answer is not a single number. What matters most is how representative the training data is of the patterns the model needs to learn and the area where the model will be applied. Important caveat: the maximum number of labels you can upload via OlmoEarth Studio is 100,000. If you have a dataset with more than 100,000 labels, please contact the OlmoEarth team for more dedicated assistance!

Is your data AI-ready? Below are some important considerations for data used in fine-tuning.

More data is (usually) better​

More data, assuming good quality, will improve model performance. Oftentimes, there are many sources of relevant data that benefit your model by adding more examples of what you hope to detect and/or other classes to help distinguish from what you want to identify.

Data quality​

As the old saying goes, "garbage in, garbage out." You should understand the quality of labels, including doing a visual inspection of the data before creating a model. If your data quality is mixed, you may consider evaluating how well your model performs with all the data vs. only higher-confidence data.

Spatial distribution​

Many points/polygons collected from a small location are often less helpful than the same number collected from a broader area, as spatial diversity helps the model generalize. Too much clustering can make some labels redundant. In general, ground truth observations should reflect the spatial heterogeneity of your area of interest, though the right balance depends on your task and deployment area.

Class distribution​

Your labeled data (ground truth or annotations) should be evenly distributed to prevent the model from overfitting to the majority classes and neglecting the minority ones. Ideally, the spatial and categorical distribution of training samples should approximate the expected distribution across your area(s) of interest.

Negative / Non-target data​

In addition to learning positive examples, models also require negative or non-target classes to produce high-quality results. A couple of examples:

  • A quality model for detecting man-made water reservoirs needs examples of non-reservoirs (e.g., lakes, rivers) to distinguish them from the desired output of man-made reservoirs.
  • If you run inference over a broad area but are only interested in certain features, you can group non-target classes into a single "other" class. For example, a crop type model may include crop classes (e.g., wheat, maize, soy) and an "other" class representing all other land cover, such as non-target crops, built-up areas, or water bodies. If inference is limited to a well-defined area (e.g., high-quality cropland extent), an "other" class may not be necessary if the training data is representative of the crops within that area.

If you do not have annotations of negative examples or other classes expected within the fine-tuning and inference area, the model may incorrectly classify unrelated features as the target class.

Volume of labeled data​

A minimum number of labels per class should reflect the area or interest's heterogeneity and the task's complexity. For example, water vs land in a small, homogenous area is simple compared to more complex or nuanced classes, like different types of trees.

Foundation models can reduce the amount of training data required to achieve reasonably accurate results; however, this should not be misconstrued to mean that extremely scarce data will produce good results.

  • For global mangrove mapping, 100k labels achieved an F1 score of 0.981, while just 10k points achieved a similar score of 0.969.
  • For crop type mapping in a 3000 km2 area, 1000 labels yielded an average F1 score of 0.79; increasing the number of labels continued to improve the F1 until all 3000 labels were used, yielding a 0.92 F1 score.

Polygon data​

Polygons should be clean, continuous shapes. Self-intersecting edges (e.g., figure-eight or bowtie shapes) are a common issue when collecting data with GPS devices or drawing polygons manually. OlmoEarth will attempt to repair these automatically, but for best results, review and clean your polygon data before uploading. Polygons made up of multiple separate parts are also supported.