This is a computer vision project for plant leaf diseases. In this project we'll be doing image dataset analysis, data augmentation, image transformations and image classification.
To setup the project, you need to launch the following command:
git clone https://github.com/arlaine4/Leaffliction && cd Leaffliction
bash setup.sh
source venv/bin/activateA program named Distribution.py is created to extract and analyze the image dataset of plant leaves. Pie charts and Bar charts are generated for each plant type, using images available in the subdirectories of the given input directory.
To balance the data set, we have a second program called Augmentation.py. It uses data augmentation techniques such as rotating, flipping, cropping, etc. To create 6 types of augmented images for each original image.
In this part, the Transformation.py program is created to directly extract features from plant leaf images. Transformations like Gaussian blur, ROI (Region of Interest) objects, object analysis, etc., are applied to images to facilitate key information extraction. For this part we use the PlantCV library.
The final step involves developing two programs: train.py and predict.py.
The train program uses augmented images to learn the characteristics of specified leaf diseases, utilizing a CNN through Keras. The learning results are saved and returned in a .zip archive.
The predict program takes a leaf image as input, displays it along with its transformations, and predicts the type of disease specified in the leaf.