Ground Truth Data

<div> <div> <div> <div>&nbsp;</div> </div> </div> </div> <div> <div> <div> <div> <div> <div> <p>The repository contains a dataset of labeled RGB and NDVI tree images with health statuses and the related trainined classification model. The dataset images are generated from the original images using geometric transformations from the Albumentations library. The labels folder contains three CSV files, each representing a health status. Health status 1 is labeled as Asymptomatic, 2 as Mild symptoms, and 3 and 4 as Evident symptomatic/compromised. We decided to merge labels 3 and 4 to create a more balanced dataset.</p> <p>The leading number "i_" indicates that the image has been generated through augmentation. This means that the original image has been modified to create this new version. The original unmodified image would be named "DJI_" without the leading number.</p> <p>For example:</p> <p>Original image: DJI_20240525124035_0032_NDVI_0.JPG</p> <p>Augmented images: 1_DJI_20240525124035_0032_NDVI_0.JPG, 2_DJI_20240525124055_0042_RGB_0.JPG, ...</p> <p>&nbsp;</p> <p>The classification model was chosen as the best-performing one among the various created during training. The model is a custom neural network designed to handle both NDVI and RGB images.</p> </div> </div> </div> </div> </div> </div>