app.py

#!/usr/bin/env python
"""
Application for ResNet50 trained on ImageNet-1K.
"""
# Standard Library Imports
import gradio as gr

# Third Party Imports
import torch
from torchvision import models

# Local Imports
from inference import inference


def load_model(model_path: str):
    """
    Load the model.
    """
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    print(f"Using device: {device}")
    
    # Initialize a fresh model without pretrained weights
    model = models.resnet50(weights=None)
    model = model.to(device)

    # Load custom weights
    state_dict = torch.load(model_path, map_location=device)
    
    # Debug: Print original state dict keys
    print("\nOriginal state dict keys:", list(state_dict['model_state_dict'].keys())[:5])
    
    # Remove the 'model.' prefix from state dict keys
    new_state_dict = {}
    for key, value in state_dict['model_state_dict'].items():
        new_key = key.replace('model.', '')
        new_state_dict[new_key] = value
    
    # Debug: Print modified state dict keys
    print("Modified state dict keys:", list(new_state_dict.keys())[:5])
    print("Model state dict keys:", list(model.state_dict().keys())[:5])
    
    # Load the modified state dict
    try:
        model.load_state_dict(new_state_dict)
        print("Successfully loaded model weights")
    except Exception as e:
        print(f"Error loading state dict: {str(e)}")
        raise e
    
    model.eval()
    return model


def load_classes():
    """
    Load the ImageNet classes
    """
    weights = models.ResNet50_Weights.IMAGENET1K_V1
    classes = weights.meta["categories"]
    print(f"Loaded {len(classes)} classes")
    return classes


def inference_wrapper(image, alpha, top_k, target_layer):
    """
    Wrapper function for inference with error handling
    """
    try:
        if image is None:
            return {"Error": 1.0}, None
            
        results = inference(
            image, 
            alpha, 
            top_k, 
            target_layer, 
            model=model, 
            classes=classes
        )
        
        if results is None:
            return {"Error": 1.0}, None
            
        return results
        
    except RuntimeError as e:
        error_msg = str(e)
        print(f"Error in inference: {error_msg}")
        
        if "out of memory" in error_msg.lower():
            return {"GPU Memory Error - Please try again": 1.0}, None
        return {"Runtime Error: " + error_msg: 1.0}, None
        
    except Exception as e:
        error_msg = str(e)
        print(f"Error in inference: {error_msg}")
        return {"Error: " + error_msg: 1.0}, None


def main():
    """
    Main function for the application.
    """
    global model, classes
    
    try:
        print(f"Gradio version: {gr.__version__}")
        
        # Load the model at startup
        model = load_model("resnet50_imagenet1k.pth")
        classes = load_classes()

        with gr.Blocks() as demo:
            gr.Markdown(
                """
                # ResNet50 trained on ImageNet-1K
                A large-scale image classification dataset with 1.2 million training images across 1,000 object categories.
                """
            )

            with gr.Tab("Predictions & GradCAM"):
                gr.Markdown(
                    """
                    View model predictions and visualize where the model is looking using GradCAM.
                    
                    ## Steps to use:
                    1. Upload an image or select one from the examples below
                    2. Adjust the sliders (optional):
                       - Activation Map Transparency: Controls the blend between original image and activation map
                       - Number of Top Predictions: How many top class predictions to show
                       - Target Layer Number: Which network layer to visualize (deeper layers show higher-level features)
                    3. Click "Generate GradCAM" to run the model
                    4. View the results:
                       - Left: Original uploaded image
                       - Right: Model predictions and GradCAM visualization showing where the model focused
                    """
                )
                
                # Define inputs
                with gr.Row():
                    img_input = gr.Image(
                        label="Input Image",
                        type="numpy",
                        height=224,
                        width=224
                    )
                    with gr.Column():
                        label_output = gr.Label(label="Predictions")
                        gradcam_output = gr.Image(
                            label="GradCAM Output",
                            height=224,
                            width=224
                        )

                with gr.Row():
                    alpha_slider = gr.Slider(
                        minimum=0,
                        maximum=1,
                        value=0.5,
                        step=0.1,
                        label="Activation Map Transparency"
                    )
                    top_k_slider = gr.Slider(
                        minimum=1,
                        maximum=10,
                        value=3,
                        step=1,
                        label="Number of Top Predictions"
                    )
                    target_layer_slider = gr.Slider(
                        minimum=1,
                        maximum=6,
                        value=4,
                        step=1,
                        label="Target Layer Number"
                    )

                gradcam_button = gr.Button("Generate GradCAM")
                
                # Set up the click event
                gradcam_button.click(
                    fn=inference_wrapper,
                    inputs=[
                        img_input,
                        alpha_slider,
                        top_k_slider,
                        target_layer_slider
                    ],
                    outputs=[
                        label_output,
                        gradcam_output
                    ]
                )

                # Examples section for Gradio 5.x
                examples = [
                    [
                        "assets/examples/cat.jpg",
                        0.5,
                        3,
                        4
                    ],
                    [
                        "assets/examples/frog.jpg",
                        0.5,
                        3,
                        4
                    ],
                    [
                        "assets/examples/bird.jpg",
                        0.5,
                        3,
                        4
                    ],
                    [
                        "assets/examples/car.jpg",
                        0.5,
                        3,
                        4
                    ],
                    [
                        "assets/examples/truck.jpg",
                        0.5,
                        3,
                        4
                    ],
                    [
                        "assets/examples/horse.jpg",
                        0.5,
                        3,
                        4
                    ],
                    [
                        "assets/examples/plane.jpg",
                        0.5,
                        3,
                        4
                    ],
                    [
                        "assets/examples/ship.png",
                        0.5,
                        3,
                        4
                    ]
                ]

                gr.Examples(
                    examples=examples,
                    inputs=[
                        img_input,
                        alpha_slider,
                        top_k_slider,
                        target_layer_slider
                    ],
                    outputs=[
                        label_output,
                        gradcam_output
                    ],
                    fn=inference_wrapper,
                    cache_examples=False,  # Disable caching to prevent memory issues
                    label="Click on any example to run GradCAM"
                )

            # Queue configuration
            demo.queue(max_size=1)  # Only allow one job at a time

            # Launch with minimal memory usage
            demo.launch(
                server_name="0.0.0.0",
                server_port=7860,
                share=True
            )
            
    except Exception as e:
        print(f"Error during startup: {str(e)}")
        if torch.cuda.is_available():
            torch.cuda.empty_cache()


if __name__ == "__main__":
    main()