# https://www.globalsino.com/ICs/
# (Text and image) contrastive learning

from sentence_transformers import SentenceTransformer, util
from PIL import Image

# Load CLIP model
model = SentenceTransformer('clip-ViT-B-32')

# Encode an image
img_emb = model.encode(Image.open(r"C:\GlobalSino20230219\ICs\images2\Animal.jpg"))

# Encode texts
text_emb = model.encode(['A person', 'A dog', 'A transitor'])

# Compute cosine similarities 
cos_scores = util.cos_sim(img_emb, text_emb)
print(cos_scores)