Quick start

HyperIA is the H2M module where you manage datasets, run trainings (classical ML models on spectral data), and inspect metrics and spectral signatures. This guide follows the product flow shown in the screenshots below. Some UI strings may appear in English or Spanish depending on your locale.

Log in with your email and password. After authentication, choose a project—datasets and trainings belong to that project.

H2M login screen with email, password, and Log in — Sign in to H2M (Hyper2Multi) with your account.

Project picker: list of projects to open — Choose the project you want to work in.

2. Open the HyperIA module

From Available modules, select HyperIA (CPU icon). Other modules (HyperLabel, HyperDrones) are documented separately.

Available modules: HyperIA, HyperDrones, and HyperLabel cards — Open HyperIA to train models on your datasets.

3. Datasets overview

Inside HyperIA, use the top breadcrumb: HyperIA → Datasets lists your datasets. You can switch to Training from the same menu when you are ready to train models.

HyperIA Datasets page with breadcrumb and search — Use the breadcrumb HyperIA → Datasets (or switch to Training).

4. Create a dataset

Create a new dataset and set a name you will recognize later (it appears in the training wizard).

New dataset dialog: name field and Create — Create a dataset and give it a clear name.

5. Add labels and CSVs

Each class in your problem needs label data:

Use + Add label to attach CSV files.
Upload one CSV per class. Each row corresponds to one pixel in the multispectral/hyperspectral image (see on-screen help).
In the New label dialog, you can drag multiple .csv files (up to 10 at a time). A label is created for each file name. All files must use the same columns in the same order, with a header row.

Dataset setup: upload CSVs per class and Add label — Upload one CSV per class—each row is one pixel. Use + Add label to add more classes.

New label modal: drag CSV files, up to 10 files — Labels are created from each file name. CSVs need headers; columns must match across files.

6. Check the dataset

Open your dataset to confirm number of bands, number of samples, and band range. The Spectral signature chart plots a random subset of rows so you can sanity-check reflectance before training.

Dataset detail: band count, samples, spectral signature chart — Review Information for band count, sample size, and a spectral signature preview.

7. Start a training

Go to HyperIA → Training to see trainings. Start New training and complete the wizard:

Name — identify the run (e.g. experiment or model variant).
Dataset — pick one of your datasets.
Configuration — choose a classifier (e.g. Support Vector Machines, Random Forest), a selector (e.g. SFS), and the initial/final band range the selector should use.

HyperIA Training section: Trainings list — Go to HyperIA → Training to start and manage trainings.

New training step 1: training name — New training wizard — step 1: name the run.

New training step 2: pick dataset from dropdown — Step 2: select which dataset to train on.

New training step 3: classifier, selector, band range — Step 3: choose classifier, selector (e.g. SFS), and initial/final band indices.

After you confirm, the app runs training with your settings.

8. Metrics and comparison

Open the project’s Metrics (or equivalent) view to compare models: evolution of accuracy, kappa, and F1 by number of bands, plus spectral visualization (and optionally confusion matrix) for each classifier–selector combination.

Metrics page: charts and spectral visualization for trained models — Open Metrics to compare runs, see accuracy, kappa, F1, and spectral signatures.

Datasets — more detail on dataset management.
Trainings — more detail on training runs and evaluation.