This repository recalls using Hyper3dnet for trees classification using the MAAP and hyperspectral data
Description
At the MAAP stack, the GPU use is enabled.
This code conduct a supervised training on Hyperspectral data from PRISMA mission using the Hyper3Dnet model.
The model is implemented with Tensoflow package and it runs using GPU on the MAAP.
The workspace
The Jupyter MAAP stack dev file sample in : dev_file_GPU.txt allows using the GPU.
This devfile contains a nodeSelector and a toleratino to be able to use the gpu node.datasets and packages
First thing to do after creating GPU workspace is to dowload data that are stored in the S3. For this task you should :
*use the S3 script (in windows) :
echo $MAAP_ENV_TYPE
export MAAP_ENV_TYPE=VAL
export CLIENT_ID=97262f0b-d3ca-4492-bcf8-9a0e12bdede8-
Downloading request using maap-s3.py Since data are stored in : maap-scientific-data/shared/imane
list first stored files in that repository by runing :maap-s3.py list /maap-scientific-data/shared/imane/Download the image file (.mat) : maap-s3.py download maap-scientific-data/shared/imane/filename path_where_to_save_data
ex : maap-s3.py download maap-scientific-data/shared/imane/france_region2_data.mat /projects/data/france_region2_data.matDownload the ground truth file (.mat) : maap-s3.py download maap-scientific-data/shared/imane/filename path_where_to_save_data
ex : maap-s3.py download maap-scientific-data/shared/imane/france_region2_GT.mat /projects/data/france_region2_GT.mat
*The second step would be installing the required packages for the code processing in your working environement weither it's the base env or a vir env
The required packages are listed in **requirements text file**
Please run :
**pip install -r requirements.txt**
In your env run : **conda install -c conda-forge psutil** Execution
Following the jupyter notebook steps leads to Training the Hyper3Dnet Model with GPU use. A hint of the cells outputs are in the notebook : Classiifcation_using_hyper3Dnet.ipynb