https://github.com/sapt36/plant_disease_recognition_using_cnn
Digital Image Processing Final Project @NTU BME
https://github.com/sapt36/plant_disease_recognition_using_cnn
cnn-classification datamining digitalimageproccessing
Last synced: 9 months ago
JSON representation
Digital Image Processing Final Project @NTU BME
- Host: GitHub
- URL: https://github.com/sapt36/plant_disease_recognition_using_cnn
- Owner: sapt36
- Created: 2025-03-27T03:56:23.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-08-18T08:58:01.000Z (10 months ago)
- Last Synced: 2025-08-18T10:28:41.813Z (10 months ago)
- Topics: cnn-classification, datamining, digitalimageproccessing
- Language: Python
- Homepage: https://www.kaggle.com/code/andrewchen111/plant-disease-recognition-using-cnn-acc-95
- Size: 12.5 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Plant Disease Classification Using Convolutional Neural Networks
### Project Introduction
This project utilizes deep learning technology and Convolutional Neural Networks (CNNs) to achieve automated classification of plant leaf diseases. The goal is to build an image classification model capable of accurately identifying three states: "Healthy", "Powdery Mildew", and "Rust".
- Two Layer CNN Architecture
- Four Layer CNN Architecture
### Core Technologies and Methods
- **Model Architecture**: Experiments were conducted with two-layer and four-layer CNN models, with parameters (such as number of layers, convolutional kernel size, activation function, Dropout rate, training epochs, and batch size) adjusted to optimize performance.
- **Dataset**: A public dataset from Kaggle was used, containing 1,472 high-resolution plant leaf images.
- **Image Preprocessing**:
- **Image Resizing**: Original images were adjusted to a suitable size for model training to improve processing efficiency.
- **Data Augmentation**: Various techniques were employed, such as random contrast adjustment, horizontal and vertical flipping, scaling, rotation, and **Gaussian noise addition**, to significantly enhance the model's generalization capabilities and accuracy.
- **Key Achievements**: High F1 scores were successfully achieved, demonstrating the potential of deep learning for plant disease classification and laying a foundation for future research.
### Project Results
- **Model Performance**: Through a carefully designed model architecture and preprocessing techniques, the model achieved high accuracy on the test set.
- **Innovation**: The addition of Gaussian noise, in particular, was proven to be a key factor in improving the model's robustness and accuracy.
### Relevant Links
- **Code and Files**: [NTU Space](https://www.space.ntu.edu.tw/navigate/a/#/s/564E5DE3D9E14F7FA912D0844D28038B6BL)
- **Kaggle Workspace**: [Project Page](https://www.kaggle.com/code/andrewchen111/plant-disease-recognition-using-cnn-acc-95)