import gradio as gr
import torch
from torch import nn
from torchvision import models, transforms
from huggingface_hub import hf_hub_download
from PIL import Image
import os
import logging
import requests
from io import BytesIO

# Setup logging
logging.basicConfig(level=logging.INFO)

# Define the number of classes
num_classes = 3

# Download model from Hugging Face
def download_model():
    model_path = hf_hub_download(repo_id="jays009/Resnet3", filename="model.pth")
    return model_path

# Load the model from Hugging Face
def load_model(model_path):
    model = models.resnet50(pretrained=False)
    num_features = model.fc.in_features
    model.fc = nn.Sequential(
        nn.Dropout(0.5),
        nn.Linear(num_features, 3)  # 3 classes
    )
    checkpoint = torch.load(model_path, map_location=torch.device("cpu"))
    model.load_state_dict(checkpoint['model_state_dict'])

    # Rename keys to match the model definition
state_dict['fc.weight'] = state_dict.pop('fc.1.weight')
state_dict['fc.bias'] = state_dict.pop('fc.1.bias')

# Load the modified state dict
model.load_state_dict(state_dict)
    model.eval()
    return model


# Path to your model
model_path = hf_hub_download(repo_id="jays009/Resnet3", filename="model.pth")
model = load_model(model_path)


# Download the model and load it
model_path = download_model()
model = load_model(model_path)

# Define the transformation for the input image
transform = transforms.Compose([
    transforms.Resize(256),
    transforms.CenterCrop(224),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])

# Prediction function for an uploaded image

def predict_from_image_url(image_url):
    try:
        # Download the image from the provided URL
        response = requests.get(image_url)
        response.raise_for_status()
        image = Image.open(BytesIO(response.content))

        # Apply transformations
        image_tensor = transform(image).unsqueeze(0)

        # Perform prediction
        with torch.no_grad():
            outputs = model(image_tensor)
            predicted_class = torch.argmax(outputs, dim=1).item()

        # Interpret the result
        if predicted_class == 0:
            return {"result": "The photo is of Fall Army Worm with problem ID 126."}
        elif predicted_class == 1:
            return {"result": "The photo shows symptoms of Phosphorus Deficiency with Problem ID 142."}
        elif predicted_class == 2:
            return {"result": "The photo shows symptoms of Bacterial Leaf Blight with Problem ID 203."}
        else:
            return {"error": "Unexpected class prediction."}

    except Exception as e:
        return {"error": str(e)}


demo = gr.Interface(
    fn=predict_from_image_url,
    inputs="text",
    outputs="json",
    title="Maize Disease Classification",
    description="Enter a URL to an image for classification (Fall Army Worm, Phosphorus Deficiency, or Bacterial Leaf Blight).",
)

if __name__ == "__main__":
    demo.launch()