app.py

import os
from huggingface_hub import login

login(token=os.getenv("HF_TOKEN"))

import gradio as gr
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision.models as models
import numpy as np
from PIL import Image
import json
from huggingface_hub import hf_hub_download

# =========================
# 📁 MODEL CONFIG
# =========================

MODEL_REPO = "InfoBayAI/resnet18-mri-anatomy-classifier"

weights_path = hf_hub_download(repo_id=MODEL_REPO, filename="pytorch_model.bin")
config_path = hf_hub_download(repo_id=MODEL_REPO, filename="config.json")
classes_path = hf_hub_download(repo_id=MODEL_REPO, filename="classes.json")

with open(config_path, "r") as f:
    config = json.load(f)

with open(classes_path, "r") as f:
    classes = json.load(f)

# =========================
# ⚙️ DEVICE
# =========================
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# =========================
# 🧠 LOAD MODEL
# =========================
model = models.resnet18(pretrained=False)

model.conv1 = nn.Conv2d(
    config["num_channels"],
    64,
    kernel_size=7,
    stride=2,
    padding=3,
    bias=False
)

model.fc = nn.Linear(model.fc.in_features, config["num_classes"])

model.load_state_dict(torch.load(weights_path, map_location=device))
model.to(device)
model.eval()

print("✅ MRI Model Loaded Successfully")

# =========================
# 🧼 PREPROCESS
# =========================
def preprocess_image(image):
    img = image.convert("L")
    img = img.resize((config["input_size"], config["input_size"]))
    img = np.array(img)

    # Normalize (same as training)
    img = (img / 255.0 - 0.5) / 0.5

    img = torch.tensor(img).unsqueeze(0).unsqueeze(0).float()
    return img.to(device)

# =========================
# 🔮 PREDICT
# =========================
def predict_mri(image): 
    if image is None:
        return "⚠️ Please upload an MRI image.", None

    img = preprocess_image(image)

    with torch.no_grad():
        output = model(img)
        probs = F.softmax(output, dim=1)

    probs_np = probs.cpu().numpy()[0]

    # Sort probabilities (highest → lowest)
    sorted_indices = np.argsort(probs_np)[::-1]

    top_label = classes[sorted_indices[0]]
    top_conf = probs_np[sorted_indices[0]] * 100

    # Format probabilities nicely
    formatted_probs = "\n".join([
        f"{classes[i]} — {probs_np[i]*100:.2f}%"
        for i in sorted_indices
    ])

    result_text = f"""
🧠 MRI Classification Result

📌 Prediction: {top_label}
🔍 Confidence: {top_conf:.2f}%

📊 Probabilities:
{formatted_probs}

⚠️ AI-assisted output — not a medical diagnosis
"""

    # For chart display
    probs_dict = {
        classes[i]: float(probs_np[i])
        for i in range(len(classes))
    }

    return result_text, probs_dict

# =========================
# 🎨 WIDE LANDSCAPE UI
# =========================
def create_interface():
    with gr.Blocks(
        theme=gr.themes.Soft(),
        css="""
        .gradio-container {
            max-width: 1600px !important;
            margin: auto;
        }

        .gr-image {
            min-height: 500px !important;
        }

        textarea {
            font-size: 16px !important;
        }

        h1, h2, h3 {
            text-align: center;
        }
        """
    ) as interface:

        gr.Markdown("# 🧠 MRI Anatomy Classifier")
        gr.Markdown("### Upload an MRI scan to identify the body region")

        with gr.Row(equal_height=True):

            # LEFT PANEL (IMAGE)
            with gr.Column(scale=1.2):
                image_input = gr.Image(
                    type="pil",
                    label="📤 Upload MRI Image",
                    height=500,
                    sources=["upload"]
                )

                with gr.Row():
                    predict_btn = gr.Button("🚀 Analyze MRI", variant="primary")
                    clear_btn = gr.Button("🧹 Clear")

            # RIGHT PANEL (RESULT - WIDER)
            with gr.Column(scale=1.8):
                output_text = gr.Textbox(
                    label="📋 Result",
                    lines=18
                )

                output_chart = gr.Label(
                    label="📊 Confidence Breakdown"
                )

        gr.Markdown("---")
        gr.Markdown("💡 Tip: Use clear MRI scans for better predictions.")

        predict_btn.click(
            fn=predict_mri,
            inputs=image_input,
            outputs=[output_text, output_chart]
        )

        clear_btn.click(
            fn=lambda: (None, "", None),
            inputs=[],
            outputs=[image_input, output_text, output_chart]
        )

    return interface

# =========================
# 🚀 LAUNCH
# =========================
if __name__ == "__main__":
    app = create_interface()
    app.launch(share=True)