app.py

import gradio as gr
from transformers import AutoFeatureExtractor, AutoModelForImageClassification
from PIL import Image
import torch

# Load model and feature extractor
model_name = "yusuf802/Leaf-Disease-Predictor"
extractor = AutoFeatureExtractor.from_pretrained(model_name)
model = AutoModelForImageClassification.from_pretrained(model_name)

# Class map (replace this with full mapping if needed)
class_map = {
    "0": "Apple_Black_rot",
    "1": "Apple_healthy",
    "33": "Tomato_Late_blight",
    "43": "Wheat_nitrogen_deficiency"
}

# Prediction function
def predict_leaf_disease(image):
    image = image.convert("RGB")
    inputs = extractor(images=image, return_tensors="pt")
    with torch.no_grad():
        outputs = model(**inputs)

    probabilities = torch.softmax(outputs.logits, dim=1)
    pred_class = torch.argmax(probabilities, dim=1).item()
    confidence = round(probabilities[0][pred_class].item() * 100, 2)
    result = class_map.get(str(pred_class), "Unknown Class")

    return f"{result} ({confidence}%)"

# Gradio Interface
demo = gr.Interface(
    fn=predict_leaf_disease,
    inputs=gr.Image(type="pil"),
    outputs="text",
    title="🌿 LeafCare - Plant Disease Predictor",
    description="Upload a leaf image to detect plant diseases using AI/ML."
)

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