# Import necessary libraries
from transformers import AutoProcessor, AutoModelForZeroShotImageClassification
from PIL import Image
import requests
import torch
import matplotlib.pyplot as plt

# Load the pre-trained model and processor
checkpoint = "openai/clip-vit-large-patch14"
model = AutoModelForZeroShotImageClassification.from_pretrained(checkpoint)
processor = AutoProcessor.from_pretrained(checkpoint)

# Load and display the image
url = "https://cdn.akamai.steamstatic.com/steam/apps/1026420/header.jpg?t=1657716289"
image = Image.open(requests.get(url, stream=True).raw)
plt.imshow(image)
plt.show()

# Specify candidate labels for zero-shot classification
candidate_labels = ["tree", "car", "bike", "cat"]

# Prepare inputs for the model
inputs = processor(text=candidate_labels, images=image, return_tensors="pt", padding=True)

# Make predictions
outputs = model(**inputs)
logits = outputs.logits_per_image  # shape: [batch_size, num_classes]
probs = logits.softmax(dim=1)  # Convert to probabilities

# Get and print the most likely class
predicted_class_idx = probs.argmax(-1).item()
predicted_class = candidate_labels[predicted_class_idx]
print(f'Predicted class: {predicted_class}')