image classification predictions

[arial133232]

I am trying to run image classification inference with HyperSIGMA on my own hyperspectral .mat data file. I have the pretrained weights and all dependencies set up.

However, I’m not sure how to:

Properly format and load my .mat file for the model.

Skip the training steps and only run the inference/prediction using the pretrained weights.

Use the provided scripts or demo notebooks to run predictions on new data.

Could you please provide a simple example or instructions on how to:

Load a new .mat hyperspectral data cube and preprocess it as expected.

Run the pretrained model to get classification predictions on this data.

Interpret/save the output predictions.

Thanks in advance for your help!

[jinyaoWHU]

@arial133232 Thank you for your question. We have already provided the complete inference code demo cls hypersigma inference.ipynb and demo seg hypersigma inference.ipynb ,you only need to modify the data loading path and set the number of classes to directly use the pretrained model for inference.
Please note that our model follows a pretraining–finetuning strategy. While it is possible to run inference directly with the pretrained weights, we strongly recommend finetuning the model on your own dataset first to achieve better performance, especially on new scenes or data domains.