app.py

import os
import torch
import torch.nn as nn
import torchvision.transforms as transforms
import timm  # EfficientNet library
import gradio as gr
from PIL import Image

# 🔹 Set device (CPU since no GPU)
device = torch.device("cpu")

# 🔹 Load EfficientNet-B0 Model
model = timm.create_model("efficientnet_b0", pretrained=False, num_classes=2)  # Binary Classification
model.load_state_dict(torch.load("efficientnet_b0_model.pth", map_location=device))  # Load trained weights
model.eval()
model.to(device)

# 🔹 Define Data Transforms (Same as Training)
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])  # Standard normalization
])

# 🔹 Paths to example images
correct_examples = [
    # "dataset/correct/correct (1).webp",
    "dataset/correct/correct (2).webp",
    # "dataset/correct/correct (3).jpeg",
    # "dataset/correct/correct (4).jpeg",
    "dataset/correct/correct (5).webp",
    # "dataset/correct/correct (6).jpeg"
]

incorrect_examples = [
    "dataset/incorrect/incorrect (1).webp",
    "dataset/incorrect/incorrect (3).webp",
    # "dataset/incorrect/incorrect (10).webp",
    # "dataset/incorrect/incorrect (12).webp",
    # "dataset/incorrect/incorrect (15).webp",
    # "dataset/incorrect/incorrect (16).webp"
]

# 🔹 Function to classify an image
def classify_installation(image):
    """Classify if the bed installation is correct or incorrect using EfficientNet-B0."""
    image = transform(image).unsqueeze(0).to(device)  # Apply transforms and add batch dimension
    
    with torch.no_grad():
        outputs = model(image)
        _, predicted = torch.max(outputs, 1)  # Get class with highest probability
    
    return "✅ Correct Installation" if predicted.item() == 0 else "❌ Incorrect Installation"

# 🔹 Function to load image from file path
def load_image(image_path):
    return Image.open(image_path).convert("RGB")

# 🔹 Function to process selected example image
def process_example(image_path):
    image = load_image(image_path)
    return classify_installation(image), image

# 🔹 Gradio UI
with gr.Blocks() as demo:
    gr.Markdown("# Installation Classifier")
    gr.Markdown("Upload an image or select one from the examples below to check if the bed installation is correct.")

    with gr.Row():
        uploaded_image = gr.Image(type="pil", label="Upload an image for testing")
        output_text = gr.Textbox(label="Result")

    gr.Markdown("### Check Installations Examples (Click Button to Classify)")
    with gr.Row():
        correct_buttons = []
        for i, img_path in enumerate(correct_examples):
            with gr.Column():
                gr.Image(value=load_image(img_path), interactive=False, width=150, height=150)
                btn = gr.Button(value="Check Accuracy")
                correct_buttons.append((btn, img_path))

    with gr.Row():
        incorrect_buttons = []
        for i, img_path in enumerate(incorrect_examples):
            with gr.Column():
                gr.Image(value=load_image(img_path), interactive=False, width=150, height=150)
                btn = gr.Button(value="Check Accuracy")
                incorrect_buttons.append((btn, img_path))

    # 🔹 Connect buttons to classification function
    for btn, img_path in correct_buttons + incorrect_buttons:
        btn.click(fn=process_example, inputs=[gr.State(img_path)], outputs=[output_text, uploaded_image])

    # 🔹 Process uploaded image
    uploaded_image.change(fn=classify_installation, inputs=[uploaded_image], outputs=output_text)

# 🔹 Launch Gradio UI
demo.launch()