CNND

This is a pytorch implementation of the paper 'Transferred Deep Learning for Anomaly Detection in Hyperspectral Imagery'. The author didn't provide their source code, so I was asked to reproduce the algorithm in this paper by my co-supervisor for his experiments.

Envirement

• pytorch 1.3
• numpy
• scipy
• sklearn

Datasets

This implementation only support dataset with a format like A.mat and A should have two key like data = A['data'] groundtruth = A['gt']/A['ground_truth'] to sotre data and labels respectively.

Here are the datasets using in this implementation:

• Training data: Salinas with 189 bands
• Testing data: AVIRIS-I.mat, AVIRIS-II.mat

Usage

You can train it directly using python train.py or test you datasets using python eval.py for the default setting. You should change the data path if you wana using other datasets.

Performance