RGB-to-NDVI-conversion dataset

Dataset for the article: Evaluating Generalization of Methods for Artificially Generating NDVI from UAV RGB Imagery in Vineyards

A full description of the article can be found at MDPI Remote Sensing: https://www.mdpi.com/2072-4292/17/3/512

As well as the base data for training Pix2Pix architectures presented on GitHub.

This data repository contains:

• p2p_data.rar which are the preprocessed chips which were used to train and evaluate the Pix2Pix models on. In this folder:
  • 256x256 and 512x512 chips each in individual folders per sensor:
    
    • ms_512 
    • ms_256
    • rgb_512
    • ms_256
  • 256x256: ms contains only training useful chips, rgb_256 contains all the test chips from can2023 and btg2021 as well
    • A=RGB
    • B=NDVI 
    • AB aligned A+B
  • 512x512: Each folder contains the structure that Pix2Pix expects:
    • A for 3band imgs, B for NDVI
    • train_A/test_A/val_A -> RGB or msRGB chips from BTG2022
    • train_B/test_B/val_B -> NDVI chips from BTG2022
    • test_2021_A -> msRGB chips from BTG2021
    • test_2023_A -> RGB chips from CAN2023
    • *_B -> NDVI chips
• orthos.rar, used in the earliest stages to create chips, and used for evaluation again later.
  
  • canyelles: RGB, ndvi, chm, dem, dsm
  • btg2021, ms, ndvi, chm, dem, dsm
  • btg2022, ms, RGB, ndvi, chm, dem, dsm
• preprocessed.rar:
  • Input files for the evaluation code, after aligning rasters, reconstructing chips to orthos