How to use 'phenoAI' python Package

(Refer PhenoAI package functioning.ipynb notebook)

The PhenoAI package can be installed from provided whl file containing the modules of the package by running the following command in an jupyter environment:

!pip install phenoAI-0.1-py3-none-any.whl

Dependencies:

opencv-python,
tensorflow,
keras,
tqdm,
segmentation_models,
xlsxwriter,
albumentations
imgaug

All the required libraries and packages will be automatically installed with installation of PhenoAI.

The PhenoAI is divided in 2 modules. Its documentation is provided below.

The `trainPhenoAI(dataset_path, epochs,learning_rate, batch_size,is_augmentation)` function is used to create a segmentation model by training it through an image dataset.

`dataset_path` : the location of the dataset containing images and labels. Optional Parameters: `epochs`, `learning_rate`, `batch_size`, `is_augmentation`(True to increase data)

This function has below sub-functions:

• `reTrain(epochs)`: For Training the model for more epochs
• `performanceReport()`: Gives us loss(or metrics) vs epochs graphs, scores of test data prediction
• `saveModel(saving_location)`: This function saves the trained phenological model.

from phenoAI import trainPhenoAI
model = trainPhenoAI (dataset_path)

Saving Image masks of Json labels...|done
Orderwise Classes names:  ['deciduous', 'coniferous']
Loading Dataset images and masks for training...|done
Training the model...
Epoch 1/35
...

model.reTrain(epochs =5)
model.saveModel(‘my_model.zip’)

for getting performance report

model.performanceReport()

Test performance:
Loss:  0.1567145701646805
iou : 0.9121432443618774
f1-score : 0.923451619338989

The function `loadModel (model, dataset_path, class_name,date_pattern)` creates an analysis object for a specific class of trees in the dataset.

The following are function's arguments:

• `model`: This is the location of the model zip file obtained from the `trainPhenoAI` module.
• `dataset_path`: This is the location of the images for analysis.
• `class_name`: This is the name of the tree category on which the analysis will be performed.
• `date_pattern`: This is used for extracting date from image name.
  This should match with image file name. It should contain 'yyyy', 'dd' and 'mm', and '*' charecter. '*' is used for covering zero aur more consequent characters.
  For Example: for file 'Sd-20221204-15_sdg.jpg', pattern can be: '*yyyymmdd-*' or '*-yyyymmdd-*'.

from phenoAI import loadModel
pheno = loadModel (model_path, dataset, vegetation_name, “_*yyyy-mmdd_*”)

Loading Vegetation segmentation Model...|done
Importing Dataset...|done
Creating ROIs in the vegetation segment...|done
Extracting Chromatic chromatic coordinate from Images...|done
Calculating phenological Parameters...|done

The analysis object created by this function includes the following functions:

• `showROIs()`: This displays the images with selected ROIs.
• `extractGCCParameters()`: This function extracts the six phenological parameters for GCC.
• `plotGCC()`: This function plots GCC versus Day of Year (DoY) graph.
• `saveGCCPlot(path_location)`: This function saves the GCC vs. DoY plot images in the
• `saveCCsTimeSeries(path_location)`: This function saves all records of GCC, BCC, RCC, and six phenological parameters in an Excel file at the given location.

Showing ROIs:

pheno.saveROIsImage("ROI_image.jpg")
pheno.showROIs()

Plotting GCC w.r.t. Day of year:

pheno.saveGCCPlot("GCC_plot.jpg")
pheno.plotGCC()

Phenological Parameters:

min, max, slope1, SoS, slope2, EoS = pheno.extractGCCParameters()

Maximum GCC: 0.4
Start of season(DoY):  117
End of season:  302
Greenary increasing rate: 0.048
Greenary decreasing rate: 0.036

Saving all the record of parameters and chromatic coordinates:

pheno.saveCCsTimeSeries(r"G:\packege testing")