Plant_disease_image_classifier

Image_model_disease (TensorFlow)

Using TensorFlow to create an image classification for plant diseases in Python.

Packages used in the Project

1. TensorFlow
2. numpy
3. matplotlib
4. fast API
5. uvicorn
6. IO
7. PIL

Initial Setup

1. Numpy package

to convert image byte data to an array

conda install numpy

2. Tensorflow package

for creating models

conda install tensorflow

3. MatPlotLib package

for visualizing model

conda install matplotlib

4. FastAPI package

for creating web API

conda install fastapi

5. uvicorn package

for creating ASGI web server implementation in Python

pip install uvicorn

6. IO package

to deal with the bytes data

pip install Python-IO

7. PIL package

for cleaning the CSV file data

pip install pillow

9. install npm in the "frontend" directory. ¹

npm install 

Important

1. If that didn't work then you can try these pip commands.
2. If that didn't work then you can try these conda commands.
3. If there is still a problem then you can install through a .whl file for that particular package.

Test Drive

1. API initialization

a. Run the file "step_6.py" in the 'api' directory.

2. Answer Fetching

a. Postman

• Open the "Postman" application.
• Select the "POST" method.
• Use this URL to send the data "http://localhost:8000/predict". (The port can be changed by you)
• Select "File" as the 'Key' and name it "file" and Select the image of your choice as 'Value' in the 'Body' label. ²
• It should display the Class of the Disease and the Confidence in the answer. like this

  {
  "class": "Corn_(maize)___Cercospora_leaf_spot Gray_leaf_spot",
  "confidence": 0.9990307092666626
  }
  

b. FastAPI docs

• Open the docs page for your API "http://localhost:8000/docs".
• Expand the "POST" tab.
• Click "Try it out" and Select the image of your choice and the "File" input. ²
• Click "Execute"
• It should display the Class of the Disease and the Confidence in the answer. (as a response) like this

  {
    "class": "Apple___Apple_scab", 
    "confidence": 0.9999998807907104
  }
  

c. Frontend (website)

• Start the website.

npm run start

• drag and drop the image of your choice ²
• It should display the Class of the Disease and the Confidence in the answer. (as a response)

Note

1. Remember to perform the initial setup before test-driving.
2. Your API program should always run when test-driving.
3. Your testing and training machine should be the same.

   NVIDIA Trained model will not run on the AMD machine

Footnotes

1. I had to do this for running the website. in the "Frontend" directory {where npm is installed} set NODE_OPTIONS=--openssl-legacy-provider. if this doesn't work then manually change "start": "react-scripts start" to "start": "react-scripts --openssl-legacy-provider start" In your "package.json" file ↩

2. I would suggest that you use an image from the "DATA/test/Apple___Apple_scab" directory 😅 The model is still weak and it cannot identify complex images 😅😅 ↩ ↩² ↩³