Crop Disease Prediction App
NOTE: This repo is Work In Progress
This app predicts the disease of the leaf image of the given plant using Deep Learning.
Technologies Used
Python, Tensorflow, FastAPI, Skimage
Workflow
- An image is uploaded to the server
- Checking is done to ensure that the image has adequate quality (brightness, contrast etc.) before it is sent for prediction. This step is performed using
image processing
- Once checked, an
AutoEncoder
model checks if the data belongs to the distribution the data is trained on or not (i.e Outlier check). - The model predicts the disease the plant.We use a Transfer Learning based classifier with
Inception v3
as backbone - In case of error, it sends an error message
Results
Classifier Results
Metric | Score |
---|---|
Accuracy | 89% |
AutoEncoder Results
Metric | Score |
---|---|
MSE | 0.05 |
MAE | 0.2 |
How to Run
- Clone the repo
- Install python requirements using
$ pip install -r requirements.txt
- Run the server using
$ python main.py