Model Overview

Satellite Data: Pixxel-TD01 and Pixxel-Firefly

The Crop Classification using Firefly model predicts pixel-wise crop labels for a selected Area of Interest (AOI) using hyperspectral imagery.

The model supports one classification variant:

Supervised Classification: Users provide reference crop information for selected fields within the same AOI. The model uses this labelled data to classify the remaining pixels within the AOI into the crop labels provided by the user.

Model Inputs

The following inputs are required:

Input Name	Description
Area of Interest	Area for which the crop classification map is required.
Labels for some fields in the AOI	Ground-truth farm GeoJSON containing reference fields labelled with a Unique ID and Crop.

Important Guidelines

Before running the model:

• Provide reference fields for all major crops present within the AOI using the reference_field GeoJSON as ground truth.
• Major crops refer to crops covering the largest area within the AOI.
• Including major crops helps the model learn dominant spectral patterns and reduces large-scale misclassification.
• Each reference field must belong to a single farm boundary.
• Each reference field must contain only one crop type.
• Do not include a reference field polygon containing multiple crops.
• If multiple crops exist within a field, split the field into separate reference field polygons for each crop type.

---

Model Output

The model generates a Crop Classification Map and corresponding classification statistics.

The Crop Classification Map is generated as a raster. Each pixel contains an integer value representing the predicted crop. The corresponding crop for each integer value can be identified using the Crop ID available in the statistics table.

Pixels with poor data quality or cloud gaps are represented using NaN values.

Statistics Table Attributes

Attribute	Description
Crop ID	ID corresponding to each predicted crop.
Pixel Count	Number of pixels corresponding to each predicted crop.
Crop Area km²	Area in km² corresponding to each predicted crop.
Percentage Coverage	Area percentage covered by each crop.
Crop	Predicted crop corresponding to the Crop ID.
Color Code	Color code assigned to each crop.

Output Files

S.No.	Output	Description	Download output as
1	Crop Classification Map	Pixel-level classification of agricultural land into different crop types for the selected AOI.	.tiff (raster format)
2	Statistics for Crop Classification Map	Derived statistics showing the total area for each predicted crop, along with model accuracy.	.csv (tabular format)

Estimated Model Run Time

*Actual time may vary.

---

Additional Details

Parameter	Details
Minimum AOI Size	No restrictions
Maximum AOI Size	5000 sq km
Geographies Supported	All geographies
Sensors Supported	Pixxel-TD01 and Pixxel-Firefly
Model Accuracy	80%

Limitations

Model accuracy is dependent on the crop's growth stage within the scene.

Model Accuracy

Model Accuracy: 80%