sits
sits copied to clipboard
e-sensing
#️⃣ Implement facilities for reproducibility and reuse of models in sits
To ensure model reproducibility and reuse, sits will save the trained model into a user defined directory (output_dir) as an RDS file.
Also, each model will be associated to a unique hash code to ensure correct continuation of a big data classification in case of failure.
Each classified tile and each block of a tile being classified by a given model will be associated to the unique hash code of this model. This will be done by including a specific metadata tag in the GeoTiff file. This tag will be named SITS_MODEL_HASH_CODE.
Temporary blocks created during a classification will be stored in a directory associated with the model hash code which is located below the output_dir selected by the user in the sits_classification() function.
