from transformers import pipeline
from PIL import Image
import os
pretrained_img_model = "nlpconnect/vit-gpt2-image-captioning"


def load_image_pipeline(img_path):
    img_path_read = Image.fromarray(img_path) 
    img_path_read.save("temp_img.jpg")
    image_to_text = pipeline("image-to-text", model=pretrained_img_model, framework="pt")
    generated_text = image_to_text("temp_img.jpg")[0]["generated_text"]
    os.remove("temp_img.jpg")
    return generated_text


if __name__=="__main__":
    imgpath = r"C:\Users\Shringar\Pictures\ar.jpg"
    img_text_generated = load_image_pipeline(imgpath)
    print(img_text_generated)