A webapp to predict the type of plant disease for a given sample. This repository is a simple project to identify different types of plant diseases. Dataset for the project was downloaded from here.
The dataset contains images of plants with different diseases. In total, there were 30+ unique categories. Some of them are
- Black Rot
- Powdery Mildew
- Late Blight
- Early Blight
- Bacterial spot
- others
Following is the list of classes which we considered for our task:
- bacterial_spot
- leaf_mold
- septoria_leaf_spot
- yellow_leaf_curl_virus
- early_blight
- mosaic_virus
- spider_mites
- leaf_blight
- powdery_mildew
- target_spot
Classes belonging to same disease but for different plants, were combined together to create more data samples. Example Peach_Bacterial_Spot, Tomato_Bacterial_Spot & Pepper_Bell_Bacterial_Spot were combined together to form a single class bacterial_spot.
Below is the model architecture
Key points:
- Number of Epochs : 50
- Activation : ReLU & SoftMax
- Optimizer : Adam
- Loss : Categorical CrossEntropy
The entire model was trained on Google Colab, were they provide a GPU environment which is preconfigured for Deep Learning purpose.
My Google Colab Notebook for this project can be found here.
- Flask
- Bokeh
- Keras
- OpenCV
- imutils
- Matplotlib
Create a local copy of this repo by cloning it
git clone https://github.com/adimyth/plant_disease.git
cd plant_disease
python main.py
- In your browser go to localhost:5000
- Select a file using the SELECT FILE button
- Click on UPLOAD
- The result for the most recent input file can be found in RESULTS section
