#from keras.models import load_model
#from fastapi import FastAPI
import gradio as gr
import requests
import numpy as np
from PIL import Image
from tensorflow import keras
from io import BytesIO
import uvicorn
#from pyModel import Item



#app = FastAPI()



#@app.post('/post')
def newPost(str):
    input_image = url_to_img(str)
    processed_image = process(input_image)
    crop = crop_predict(processed_image)
    disease = disease_predict(processed_image, crop)
    return [crop,disease]
  

def url_to_img(image_url):

    # Load the image from the URL
    response = requests.get(image_url)
    input_image = Image.open(BytesIO(response.content))
    return input_image


def process(input_image):

    # Preprocess the image to match model's input dimensions
    input_image = input_image.resize((224, 224))  # Adjust the dimensions
    input_image = np.array(input_image) / 255.0  # Normalize pixel values to [0, 1]
    processed_image = np.expand_dims(input_image, axis=0)  # Add batch dimension
    return processed_image

def crop_predict(processed_image):

    # Load the trained model
    model = keras.models.load_model('./crop_model.h5')
    # Make predictions
    predictions = model.predict(processed_image)
    # Process the predictions (e.g., get class labels)
    class_labels = ["Corn", "Potato", "Rice", "Wheat"] 
    predicted_class = np.argmax(predictions, axis=1)
    predicted_label = class_labels[predicted_class[0]]
    return predicted_label

def disease_predict(processed_image, crop):

    if crop == "Potato":
        # Load the trained model
        model = keras.models.load_model('./potato_model.h5')
        # Process the predictions (e.g., get class labels)
        class_labels = ["Potato_Early_Blight", "Potato_Healthy", "Potato_Late_Blight"] 
    
    elif crop == "Corn":
        # Load the trained model
        model = keras.models.load_model('./corn_model.h5')
        # Process the predictions (e.g., get class labels)
        class_labels = ["Corn_Common_Rust", "Corn_Gray_Leaf_Spot", "Corn_Healthy", "Corn_Northern_Leaf_Blight"] 

    elif crop == "Rice":
        # Load the trained model
        model = keras.models.load_model('./rice_model.h5')
        # Process the predictions (e.g., get class labels)
        class_labels = ["Rice_Brown_Spot", "Rice_Healthy", "Rice_Leaf_Blast", "Rice_Neck_Blast"] 

    elif crop == "Wheat":
        # Load the trained model
        model = keras.models.load_model('./wheat_model.h5')
        # Process the predictions (e.g., get class labels)
        class_labels = ["Wheat_Brown_Rust", "Wheat_Healthy", "Wheat_Yellow_Rust"]

    # Make predictions
    predictions = model.predict(processed_image)
    predicted_class = np.argmax(predictions, axis=1)
    predicted_label = class_labels[predicted_class[0]]
    return predicted_label


iface = gr.Interface(fn = newPost , inputs=['text'] , outputs=['text'])
iface.launch()