pip install -r requirements.txt
├── datasets
│ └── ip
│ ├── Indian_pines_corrected.mat
│ └── Indian_pines_gt.mat
└── pretrained
└── sstn
└── sstn_ip.pth
Calculating non-conformity scores with SACP
python main.py --model sstn --data_name ip --alpha 0.05 --base_score APS
with the following arguments:
-
model: the name of the model, including
cnn1d, cnn3d, hybrid, sstn. -
data_name: the name of dataset, including
ip, pu, sa. -
alpha: the user-specified error rate.
-
base_score: the standard non-coformity score, incuding
APS, RAPS, SAPS.
If you find this work useful for your research, please cite:
@article{liu2024spatial,
title={Spatial-Aware Conformal Prediction for Trustworthy Hyperspectral Image Classification},
author={Liu, Kangdao and Sun, Tianhao and Zeng, Hao and Zhang, Yongshan and Pun, Chi-Man and Vong, Chi-Man},
journal={arXiv preprint arXiv:2409.01236},
year={2024}
}