app.py

import gradio as gr
import numpy as np
import json
import os
from PIL import Image
import onnxruntime as ort

# ── Load ONNX model — only ~200MB RAM (vs 2GB for TensorFlow) ─
MODEL_PATH = "eye_disease_model.onnx"
CLASS_PATH  = "class_indices.json"

if not os.path.exists(MODEL_PATH):
    raise FileNotFoundError(f"Model not found: {MODEL_PATH}")

session    = ort.InferenceSession(MODEL_PATH, providers=['CPUExecutionProvider'])
input_name = session.get_inputs()[0].name

with open(CLASS_PATH) as f:
    idx_map = json.load(f)

IMG_SIZE = 224

LABEL_MAP = {
    "cataract":             "🔵 Cataract",
    "diabetic_retinopathy": "🔴 Diabetic Retinopathy",
    "glaucoma":             "🟡 Glaucoma",
    "normal":               "🟢 Normal"
}

INFO_MAP = {
    "cataract":             "Clouding of the eye lens. Treatable with surgery.",
    "diabetic_retinopathy": "Retinal damage caused by diabetes. Early detection prevents blindness.",
    "glaucoma":             "Optic nerve damage from high eye pressure. Can cause permanent vision loss.",
    "normal":               "No disease detected. Keep up with regular eye checkups!"
}

def predict(image):
    if image is None:
        return "⚠️ Please upload a retinal image."

    img  = Image.fromarray(image).convert("RGB").resize((IMG_SIZE, IMG_SIZE))
    arr  = np.array(img, dtype=np.float32) / 255.0
    arr  = np.expand_dims(arr, axis=0)           # (1, 224, 224, 3)

    preds   = session.run(None, {input_name: arr})[0][0]
    top_idx = int(np.argmax(preds))
    top_cls = idx_map[str(top_idx)]
    conf    = float(preds[top_idx]) * 100

    result  = f"## {LABEL_MAP[top_cls]}\n"
    result += f"**Confidence: {conf:.1f}%**\n\n"
    result += f"📋 {INFO_MAP[top_cls]}\n\n"
    result += "---\n### All Probabilities\n\n"
    for i in range(len(preds)):
        cls  = idx_map[str(i)]
        prob = float(preds[i]) * 100
        bar  = "█" * int(prob / 5) + "░" * (20 - int(prob / 5))
        result += f"`{LABEL_MAP[cls]:<32}` {bar} {prob:.1f}%\n\n"
    result += "\n---\n⚠️ *For educational purposes only. Always consult an ophthalmologist.*"

    return result

with gr.Blocks(title="👁️ Eye Disease Detector", theme=gr.themes.Soft()) as demo:
    gr.Markdown("# 👁️ Eye Disease Detection")
    gr.Markdown(
        "Upload a **retinal fundus image** — detects: "
        "Cataract · Diabetic Retinopathy · Glaucoma · Normal\n\n"
        "*MobileNetV2 Transfer Learning — No TensorFlow needed at runtime*"
    )
    with gr.Row():
        with gr.Column():
            img_input = gr.Image(label="Upload Retinal Image", type="numpy")
            btn = gr.Button("🔍 Analyse", variant="primary")
        with gr.Column():
            output = gr.Markdown()

    btn.click(fn=predict, inputs=img_input, outputs=output)
    img_input.change(fn=predict, inputs=img_input, outputs=output)

demo.launch()