app.py

import gradio as gr
import numpy as np
from PIL import Image
from tensorflow.keras import models
from tensorflow.keras.preprocessing.image import img_to_array
import matplotlib.pyplot as plt
import io
import base64


def inference(image, model_choice):
    label_map = {'cassava-healthy': 0, 'cassava-not-healthy:bacteria blight': 1}
    inverse_map = {v: k for k, v in label_map.items()}

    image = Image.fromarray(image.astype('uint8'), 'RGB')
    image = image.resize((64, 64))
    image_arr = img_to_array(image)
    image_arr /= 255
    image_arr = image_arr[np.newaxis, :]

    if model_choice == "Cassava Model 🍃":
        model = models.load_model("cassava_model.keras")
    else:
        model = models.load_model("/content/maize_model.keras")
    proba = model.predict(image_arr)
    label = (proba > 0.5).squeeze().astype(int)

    result = {
        "label": inverse_map.get(int(label)),
        "probability": float(proba.squeeze())
    }

    # Create visualization
    fig, ax = plt.subplots(figsize=(8, 6))
    ax.bar(['Healthy 🌿', 'Not Healthy 🍂'], [1 - result['probability'], result['probability']], color=['#2ecc71', '#e74c3c'])
    ax.set_ylim(0, 1)
    ax.set_ylabel('Probability')
    ax.set_title('Plant Health Prediction 🔍', fontsize=16, fontweight='bold')
    ax.spines['top'].set_visible(False)
    ax.spines['right'].set_visible(False)
    plt.tight_layout()

    # Convert plot to image
    buf = io.BytesIO()
    plt.savefig(buf, format='png')
    buf.seek(0)
    plot_image = Image.open(buf)

    return result["label"], f"{result['probability']:.2%} of illness(bacteria blight)", plot_image

# Custom CSS for styling
custom_css = """
#component-0 {
    max-width: 730px;
    margin: auto;
    padding: 1.5rem;
    border-radius: 10px;
    background: linear-gradient(135deg, #f6d365 0%, #fda085 100%);
    box-shadow: 0 10px 20px rgba(0,0,0,0.19), 0 6px 6px rgba(0,0,0,0.23);
}
#component-1 {
    border-radius: 10px;
    overflow: hidden;
}
#component-5 {
    border-radius: 10px;
    overflow: hidden;
    box-shadow: 0 4px 6px rgba(0,0,0,0.1);
}
.label {
    font-size: 18px !important;
    color: #2c3e50;
    font-weight: bold;
}
.output-class {
    font-size: 24px !important;
    color: #2980b9;
    font-weight: bold;
}
.output-prob {
    font-size: 20px !important;
    color: #16a085;
}
"""

# Gradio interface
with gr.Blocks(css=custom_css) as demo:
    gr.Markdown("# 🌱 Crop Diseases Detector 🕵️‍♂️")
    gr.Markdown("Upload an image of a cassava plant and let's check its health!")

    with gr.Row():
        input_image = gr.Image(type="numpy", label="📸 Upload or Capture Image")
        output_image = gr.Image(type="pil", label="🖼️ Health Prediction Visualization")

    model_choice = gr.Dropdown(["Cassava Model 🍃"], label="🤖 Select Model", value="Cassava Model 🍃")

    with gr.Row():
        detect_btn = gr.Button("🔍 Detect Plant Health", variant="primary")

    output_label = gr.Textbox(label="🏷️ Diagnosis")
    output_confidence = gr.Textbox(label="📊 Confidence")

    detect_btn.click(
        inference,
        inputs=[input_image, model_choice],
        outputs=[output_label, output_confidence, output_image]
    )

    gr.Markdown("## How to use:")
    gr.Markdown("1. 📤 Upload an image or 📸 take a picture of a cassava plant")
    gr.Markdown("2. 🤖 Select the model you want to use")
    gr.Markdown("3. 🔍 Click 'Detect Plant Health' to get the results")
    gr.Markdown("4. 📊 View the diagnosis, confidence score, and health prediction chart")

demo.launch(debug=True)