- MCNN-CP: Hyperspectral Image Classification Using Mixed Convolutions and Covariance Pooling (TGARS 2021) paper and source_code
- Oct-MCNN-HS: 3D Octave and 2D Vanilla Mixed Convolutional Neural Network for Hyperspectral Image Classification With Limited Samples (Submitted)
This code has been tested on on a personal laptop with Intel i7-9750H 2.6-GHz processor, 32-GB RAM, and an NVIDIA GTX1650 graphic card, Python 3.6, tensorflow_gpu-1.14.0, Keras-2.2.4, CUDA 10.0, cuDNN 7.6. Please install related libraries before running this code:
pip install -r requirements.txt
- IP: Indian Pines corrected and Indian Pines gt
- UH: University of Houston 2018
- UP: University of Pavia corrected and University of Pavia gt
- SA: Salinas Scene corrected and Salinas Scene gt
and put them into data directory.
- models code:
and put them into models directory.
- IP: Indian Pines code:
- UH: University of Houston code:
- UP: University of Pavia code:
- SA: Salinas Scene code:
and put them into pretrained_models directory.
python validate.py
--dataset IP # dataset_name
--model Oct-MCNN-HS # model_name
--ratio 0.99 # test_ratio
The testing result will be saved in the classification_report.txt.