app.py

import json
import torch
import urllib
import gradio as gr
from PIL import Image
import torchvision.models as models
import torchvision.transforms as transforms

# Load the JSON file with the ImageNet class labels
imagenet_labels_json_url = 'https://storage.googleapis.com/download.tensorflow.org/data/imagenet_class_index.json'
imagenet_labels_json = urllib.request.urlopen(imagenet_labels_json_url)

# Decode JSON to a Python dictionary
imagenet_labels_dict = json.load(imagenet_labels_json)

# Convert the dictionary to a list of class labels
imagenet_labels = [imagenet_labels_dict[str(i)][1] for i in range(len(imagenet_labels_dict))]


# Load a pre-trained model (for example, ResNet18)
model = models.resnet18(pretrained=True)
model.eval()

# Define the image transformation
transform = transforms.Compose([
    transforms.Resize(256),
    transforms.CenterCrop(224),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])

# Define the classification function
def classify_image(image):
    image = Image.fromarray(image.astype('uint8'), 'RGB')
    image = transform(image).unsqueeze(0)
    with torch.no_grad():
        preds = model(image)
        # Get the index and the confidence of the predicted class
    confidences = torch.nn.functional.softmax(preds, dim=1)[0] * 100
    pred_index = confidences.argmax().item()
    confidence = confidences[pred_index].item()

    # Get the human-readable class name
    pred_class_name = imagenet_labels[pred_index]

    # Determine the bar color based on the confidence
    bar_color = "#FF0000" if confidence < 50 else "#00FF00"
    
    # Create a progress bar using HTML based on the confidence
    progress_bar_html = f"""
    <div style="margin: 10px 0;">
      <label>Confidence Level:</label>
      <div style="width: 100%; background-color: #e0e0de;">
        <div style="width: {confidence}%; background-color: {bar_color}; padding: 10px; color: white; text-align: center;">
          {confidence:.2f}%
        </div>
      </div>
    </div>
    """

    # Return the class name and the progress bar HTML
    return pred_class_name, progress_bar_html

examples = [
    "examples/red_pandas.jpg",
    "examples/turtle.jpg",
    "examples/gazelle.jpg",
    "examples/cat_dog.jpg"
]

# Set up the Gradio interface
iface = gr.Interface(
    fn=classify_image,
    inputs=gr.Image(),
    outputs=[
        gr.Text(label="Class Name"),
        gr.HTML(label="Confidence Level"),
    ],
     examples=examples
)


# Launch the application
iface.launch(share=True)