To develop an ensemble model that predicts sugarcane diseases using various deep learning architectures.
Sugarcane Leaf Disease Dataset
Refer to the notebook in the /models directory.
This project focuses on predicting diseases in sugarcane using an ensemble model that combines several well-known architectures: MobileNet, InceptionV3, VGG16, Conv2D, and ResNet152.
- Requirement: With the rise of agricultural challenges, timely diagnosis of plant diseases is crucial for ensuring crop health and yield.
- Necessity: This model provides farmers with a tool to quickly identify disease types, enabling timely intervention.
- Benefits: It helps in maintaining crop health, increasing agricultural productivity, and reducing economic losses due to diseases.
- Approach: The project began with data collection and exploration, followed by model selection based on their individual performance metrics.
- MobileNet: A lightweight model suitable for mobile applications, known for its efficiency and speed.
- InceptionV3: A deep convolutional network designed to improve accuracy by utilizing parallel convolutional layers.
- VGG16: A widely used model recognized for its simplicity and deep architecture.
- Conv2D: Custom convolutional layers designed for feature extraction specific to sugarcane diseases.
- ResNet152: A deep network that uses skip connections to combat the vanishing gradient problem, enhancing learning.
- Initial data exploration and understanding.
- Data cleaning and preprocessing.
- Feature engineering and selection.
- Model training and evaluation.
- Model optimization and fine-tuning.
- Validation and testing.
- pandas
- numpy
- scikit-learn
- matplotlib
- seaborn
- tensorflow
- keras
| Model | Accuracy |
|---|---|
| MobileNet | 83.20% |
| VGG16 | 79.76% |
| Conv2D | 86.06% |
| ResNet152 | 81.21% |
| InceptionV3 | 86.06% |
| Ensemble | 90.94% |
- Application in Agriculture: Farmers can use this model to diagnose sugarcane diseases quickly, ensuring better crop management.
- Research and Development: Agricultural researchers can utilize this model to study disease patterns and develop better disease-resistant sugarcane varieties.
- Prepare the data pipeline to ensure real-time data feeding.
- Deploy the model using appropriate tools (e.g., Flask, Docker) to create an accessible web application.
- Monitor and maintain the model in production to ensure continued accuracy and performance.
- Feature 1: Integration of real-time data input for immediate predictions; not implemented due to time constraints.
- Feature 2: Development of a user-friendly mobile application; not implemented due to lack of mobile development experience.
Vivek Prakash