The repository contains code refered to the work:

Giuseppina Andresini, Annalisa Appice, Dino Ienco, Vito Recchia

DIAMANTE: A data-centric semantic segmentation approach to map tree dieback induced by bark beetle infestations via satellite images

Please cite our work if you find it useful for your research and work.

 @article{,
 title = {DIAMANTE: A data-centric semantic segmentation approach to map tree dieback induced by bark beetle infestations via satellite images},
journal = {},
volume = {},
pages = {},
year = {},
issn = {},
doi = {},
url = {},
author = {}}

The code relies on the following python3.7+ libs. Packages needed are:

• Tensorflow 2.4.0
• Pandas 1.3.2
• Numpy 1.19.5
• Matplotlib 3.4.2
• Hyperopt 0.2.5
• Keras 2.4.3
• Scikit-learn 0.24.2

The following DATASETS. The datasets used are:

The models used for the experimental setting are reported in MODELS

To launch the dataset pipeline download on your own you have to launch:

python Data/planetary_download.py

You have also to specify these parameters into planetary_download.py main() function:

• base_out_path: defines the output path, where the Sentinel 1 and Sentinel 2 images will be downlaoded
• base_geojson_path: the path where the geojson defining each scene is stored
• start_date and end_date: defining the search interval for Sentinel 2 search images.

All the required packages needed to run the download pipeline are described in Data/requirements.txt file.

The repository contains the following scripts under the folder Model:

• main.py: script to execute DIAMANTE

To replicate experiments reported in the work, you can use models and datasets stored in homonym folders. Global variables are stored in CONFIG.conf file

[Sentinel2]
pathModels = ../Models/Sentinel_2/
pathDatasetTrain = ../DS/SWIFT/planetary/sentinel_2/Train/
pathDatasetTest = ../Datasets/Sentinel_2/
pathDatasetTrainM = ../DS/SWIFT/planetary/Masks/Train/
pathDatasetTestM = ../Datasets/Masks/
nameModel=unet_resize_1_model.h5
#this is to remove channel 13 (SCL) to the tiff files
sizetest=266
resizeChannel=1
pathModelPretrained=Models/
shape=32
channels=12
channels1=0
tilesSize=32
attack=1
tiles=1




[setting]
ATTACK_LABEL = 1
PREPROCESSING=1
PREPROCESSING_MASKS=1
# if 0 the U-Net is trained
LOAD_NN = -1
#1 to perform experiment with middle and late configurations
TRAIN_LATEUNET= 0
RESIZE=1
#1 to perform prediction with trained model specified in the pathModels
PREDICTION=1
#if 1 perform prediction with middle and late unet otherwise it is performed with early stage
PREDICT_LATEUNET=0
#if 1 the late U-net is used otherwise hybrid for train and prediction
LATE=0
# if 1 the training and prediciton is performed with SUM operator fusion otherwise it is odne with CONC operator
SUM=1