This repository contains code for the Spectroscopic Transformer, a pixel-wise deep learning model for spectral tasks.
Subdirectories of this repository contain research code for specific publications and/or applications.
Installation of these packages require an existing installation of GDAL.
- libgdal needs to be version 3.9.3:
conda install conda-forge::libgdal==3.9.3
This package includes both an importable Python package and a CLI interface for model deployment and training.
Check out the docs here!
J. H. Lee, M. Kiper, D. R. Thompson, & P.G. Brodrick, Proc. Natl. Acad. Sci. U.S.A. 122 (27) e2502903122, https://doi.org/10.1073/pnas.2502903122 (2025).
For details see the cloud directory
This section is to provide a list of all of the envrionment variables related to this package and how they impact the functional behavior.
* -> replace with the uppercase name of the parameter and any dashes - to be underscores _. (i.e. --arch-proj-dim -> SPECTF_TRAIN_ARCH_PROJ_DIM)
| Evr Var Name | Description |
|---|---|
| SPECTF_TRAIN_* | Set to override the default parameters values of the SpecTf model training function. (See spectf-cloud train -h for list of parameters and their defaults) |
| SPECTF_DEPLOY_* | Set to override the default parameters values of the SpecTf model deploy function. (See spectf-cloud deploy -h for list of parameters and their defaults) |
| SPECTF_EVAL_L2A_* | Set to override the default parameters values of the L2A evaluation function. (See spectf-cloud cloud-eval l2a -h for list of parameters and their defaults) |
| SPECTF_EVAL_SPECTF_* | Set to override the default parameters values of the SpecTf evaluation function. (See spectf-cloud cloud-eval spectf -h for list of parameters and their defaults) |
| RESNET_TRAIN_* | Set to override the default parameters values of the ResNet training function. (See spectf-cloud train-comparison resnet -h for list of parameters and their defaults) |
| XGBOOST_TRAIN_* | Set to override the default parameters values of the XGBoost training function. (See spectf-cloud train-comparison xgboost -h for list of parameters and their defaults) |
