app.py

import torch
import gradio as gr
import torch.nn as nn
import torch.nn.functional as F
import os
from pathlib import Path
from torch.nn import init
import torchvision.transforms as transforms
from PIL import Image

# ... [Keep all your existing model definitions and initialization code] ...

# Precompute example image paths
example_dir = "examples"
example_images = [os.path.join(example_dir, f) for f in os.listdir(example_dir) 
                 if f.lower().endswith(('.png', '.jpg', '.jpeg', '.webp'))]

# Custom CSS for styling
css = """
.centered-examples {
    margin: 0 auto !important;
    justify-content: center !important;
    gap: 8px !important;
}
.centered-examples .thumb {
    height: 100px !important;
    width: 100px !important;
    object-fit: cover !important;
}
"""

def predict(img):
    """Process single image and return prediction"""
    if img is None:
        return "Please select or upload an image"
    
    try:
        image = Image.open(img).convert('RGB')
        tensor = preprocess(image).unsqueeze(0)  # Add batch dimension
        
        with torch.inference_mode():
            outputs = model(tensor)
            _, pred = torch.max(outputs, 1)
        
        return classes[pred.item()]
    
    except Exception as e:
        return f"Error: {str(e)}"

with gr.Blocks(title="Animal Classifier", css=css) as demo:
    gr.Markdown("## 🐾 Animal Classifier")
    gr.Markdown("Click example images below or upload your own")
    
    with gr.Row():
        input_image = gr.Image(type="filepath", label="Selected Image")
        output_label = gr.Label(label="Prediction")
    
    # Example gallery with click handling
    with gr.Row(variant="panel"):
        examples_gallery = gr.Gallery(
            value=example_images,
            label="Example Images (Click to Predict)",
            columns=7,
            height=120,
            allow_preview=False,
            elem_classes=["centered-examples"]
        )
    
    # Handle example image clicks
    def select_example(evt: gr.SelectData):
        return example_images[evt.index]
    
    examples_gallery.select(
        fn=select_example,
        outputs=input_image,
        show_progress=False
    )
    
    # Handle predictions for both upload and example clicks
    input_image.change(
        fn=predict,
        inputs=input_image,
        outputs=output_label
    )

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