This Machine Learning project, done in collaboration with Pierrickjay, aims to test the capabilities of Convolutional Neural Networks (CNNs) in the field of computer vision.
The primary goal is to train a program to accurately recognize and distinguish between images of healthy and diseased leaves.
The dataset for this project can be downloaded from here: leaves.zip
- Distribution: This part of the program helps to analyze the dataset by visualizing the distribution of images available for training.
- Augmentation: If the dataset is unbalanced, it's crucial to increase the amount of data available by creating modified images.
- Balance: This program automatically creates augmented images for the categories with the least amount of data. Once this program is run, the dataset should be balanced.
- Training: This part of the program uses transfer learning from a TensorFlow CNN and fine-tunes it with our dataset.
- Predict: This function makes a prediction about the category of a given image.
- Clean: This function removes the augmented images from the dataset.
In addition to the original program, which can be run from the backend_fastAPI/main.py file, I've also created a front-end and an API to make the project more interactive and user-friendly.
The API is created using FastAPI, and the front-end is designed using Streamlit.
The original program can be executed using the main.py file located in the backend_fastAPI directory. The program uses various command-line arguments to manage its functionalities:
- --distribution or -d: Launches the Distribution part of the program.
- --augmentation or -a: Launches the Augmentation part of the program.
- --balance or -b: Launches the Balance part of the program.
- --train or -t: Launches the Training part of the program.
- --predict or -p: Launches the Prediction part of the program.
- --clean or -c: Cleans the directory by removing the augmented images.
- --load or -l: Loads an existing model.
The --load option is necessary for the Train functionality if we wish to continue training an existing model.
Distribution
Augmentation
Prediction
The configuration options are available in the backend_fastAPI/configs/CFG.py file.
To launch the full version of the project, which includes both the front-end and the back-end, use the following command:
make
Then, open your web browser and navigate to http://localhost:8501/ to start using the project.
