Quick start

Create a raw dataset, upload .mat cubes, define labels, annotate, deploy to HyperIA, and train.

HyperLabel is where you import hyperspectral cubes, define classes, draw regions (polygons or boxes), and deploy a dataset into HyperIA for training. This guide follows the in-app flow. Some labels appear in English or Spanish depending on locale.

Note: ENVI import is planned for a future release; today HyperLabel works with .mat cubes as described below.

1. Raw dataset

Under HyperLabel → Raw Datasets, list and search datasets, then Create a new raw dataset or open an existing one.

HyperLabel Raw Datasets list with search and Create — Open HyperLabel → Raw Datasets. Create a new raw dataset or edit an existing one.

Give the dataset a name and confirm creation.

New raw dataset dialog: name and Create — Name the raw dataset and create it.

2. Upload `.mat` data

Open Upload Data in the sidebar. Upload MATLAB .mat files containing 3D cubes (height × width × bands). You can add up to 100 files per batch. Set Previsualization band (1-based index) for the grayscale preview band.

Upload Data: drag .mat files, previsualization band, up to 100 files — Upload Data: add .mat hyperspectral cubes (H×W×bands). Set the 1-based previsualization band for the viewer.

3. Labels (classes)

In Labels, create one entry per class: name, color, and optional order. Use Create / Edit from the table.

Labels table: ID, name, color, Edit — Define labels (classes): name and color for each class used in annotation.

New label form: name, color hex, optional order — Create a label with a display name, color, and optional sort order.

4. Annotate

Open an image from the dataset. Use the Band Index slider to browse spectral bands. Use the drawing tools (polygon, rectangle, etc.) to outline regions and assign a class to each shape. Save when you are done.

Annotation canvas: band index slider, polygon tools, shapes on image — Open an image, move the Band Index slider, and draw polygons or boxes per class.

Polygons and bounding regions on aerial image — Refine shapes: multiple polygons, vertices, and class colors on the canvas.

Save button and Labels panel with class assignment — Assign each shape to a class in the Labels panel, then Save.

5. Deploy to HyperIA

Go to Deploy. When nothing is deployed yet, use Deploy to build a HyperIA dataset from this raw dataset. After success, you will see the deployed row (name, band range, labels, samples).