clip_component.py

import cv2
import torch
import os
from PIL import Image
import clip

def get_token_from_clip(image):

    text_inputs = ["Bacon", "Bread", "Fruit", "Beans and Rice", "fries", "Lasagna"]
    text_tokens = clip.tokenize(text_inputs)

    device = "cpu"
    model, preprocess = clip.load("ViT-B/32")

    print("device: ", device)

    text_features = model.encode_text(text_tokens).float()
    text_features /= text_features.norm(dim=-1, keepdim=True)

    image_pil = Image.fromarray(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
    image_input = preprocess(image).unsqueeze(0).to(device)  # Add batch dimension

    with torch.no_grad():
        image_feature = model.encode_image(image_input)
        image_feature /= image_feature.norm(dim=-1, keepdim=True)

    with torch.no_grad():
        similarity = text_features.cpu().numpy() @ image_feature.cpu().numpy().T

    best_similarity = 0
    best_text_input = ""

    for i in range(similarity.shape[0]):
        similarity_num = (100.0 * similarity[i][0])
        if similarity_num > best_similarity:
            best_similarity = similarity_num
            best_text_input = text_inputs[i]

    # Print the caption for the image
    print("Best caption for the image: ", best_text_input)
    return best_text_input