This repository presents a comparative analysis of tiny machine learning models for maize crop disease identification. The project focuses on leveraging various lightweight models to accurately classify different diseases affecting maize crops.
The Datasets directory contains the "MaizeDiseaseImage_Classification_Dataset_Ghana," which includes images of maize crops affected by various diseases. The dataset is organized into five directories:
- Blight
- Common_rust
- Grapy_leaf_spot
- Healthy
- Streak_virus
Each directory contains images corresponding to the respective maize crop diseases
The models directory encompasses six subdirectories, each dedicated to a specific tiny machine learning model:
- dnn
- efficientnet
- inception
- mobilenet
- shufflenet
- squeezenet Within each model directory, you'll find the trained models and TensorFlow Lite (TFLite) files.
- DNN: Represents a basic deep neural network. EfficientNet: Utilizes the EfficientNet architecture for efficient image classification.
- Inception: Implements the Inception architecture known for its utilization of inception modules.
- MobileNet: Employs the MobileNet architecture optimized for mobile and embedded vision applications.
- ShuffleNet: Utilizes the ShuffleNet architecture designed for efficient model inference.
- SqueezeNet: Implements the SqueezeNet architecture, emphasizing smaller model size and faster inference.
Dataset Preparation: Ensure the "MaizeDiseaseImage_Classification_Dataset_Ghana" dataset is properly organized in the Datasets directory.
Model Selection: Choose the desired tiny machine learning model from the models directory based on your requirements for model size, inference speed, and accuracy.
Model Training: Train the selected model using the provided dataset or your custom dataset, depending on your specific maize crop disease identification requirements.
Model Evaluation: Evaluate the trained model's performance using appropriate evaluation metrics, such as accuracy, precision, recall, and F1 score, to assess its effectiveness in identifying maize crop diseases.
Model Deployment: Deploy the trained model to your desired platform or device for real-time maize crop disease identification and monitoring applications.
Contributions to this project are welcome. If you have any improvements, bug fixes, or additional models to contribute, feel free to submit a pull request.