How to calculate image similarity of given 2 images by using open AI Clip model - which method / AI model is best for calculating image similarity?

Question

I have prepared a small example code but It is throwing error. Can't solve the problem because it is supposed to work.

Also do you think are there any better approaches to calculate image similarity? I want to find similar cloth images. e.g. I will give an image of a coat and I want to find similar coats.

also would this code handle all dimensions of images and all types of images?

here the code

import torch
import torchvision.transforms as transforms
import urllib.request
from transformers import CLIPProcessor, CLIPModel, CLIPTokenizer
from PIL import Image

# Load the CLIP model
device = "cuda" if torch.cuda.is_available() else "cpu"
model_ID = "openai/clip-vit-base-patch32"
model = CLIPModel.from_pretrained(model_ID).to(device)

preprocess = CLIPProcessor.from_pretrained(model_ID)


# Define a function to load an image and preprocess it for CLIP
def load_and_preprocess_image(image_path):
    # Load the image from the specified path
    image = Image.open(image_path)

    # Apply the CLIP preprocessing to the image
    image = preprocess(image).unsqueeze(0).to(device)

    # Return the preprocessed image
    return image

# Load the two images and preprocess them for CLIP
image_a = load_and_preprocess_image('/content/a.png')
image_b = load_and_preprocess_image('/content/b.png')

# Calculate the embeddings for the images using the CLIP model
with torch.no_grad():
    embedding_a = model.encode_image(image_a)
    embedding_b = model.encode_image(image_b)

# Calculate the cosine similarity between the embeddings
similarity_score = torch.nn.functional.cosine_similarity(embedding_a, embedding_b)

# Print the similarity score
print('Similarity score:', similarity_score.item())

here the error message

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
[<ipython-input-24-e95a926e1bc8>](https://localhost:8080/#) in <module>
     25 
     26 # Load the two images and preprocess them for CLIP
---> 27 image_a = load_and_preprocess_image('/content/a.png')
     28 image_b = load_and_preprocess_image('/content/b.png')
     29 

3 frames
[/usr/local/lib/python3.9/dist-packages/transformers/tokenization_utils_base.py](https://localhost:8080/#) in _call_one(self, text, text_pair, add_special_tokens, padding, truncation, max_length, stride, is_split_into_words, pad_to_multiple_of, return_tensors, return_token_type_ids, return_attention_mask, return_overflowing_tokens, return_special_tokens_mask, return_offsets_mapping, return_length, verbose, **kwargs)
   2579 
   2580         if not _is_valid_text_input(text):
-> 2581             raise ValueError(
   2582                 "text input must of type `str` (single example), `List[str]` (batch or single pretokenized example) "
   2583                 "or `List[List[str]]` (batch of pretokenized examples)."

ValueError: text input must of type `str` (single example), `List[str]` (batch or single pretokenized example) or `List[List[str]]` (batch of pretokenized examples)

score 2 · Accepted Answer · answered Mar 11 '23 at 20:58

I am not sure why this code is supposed to work since it contains several errors (CLIPModel has no encode_image. CLIPProcessor.__call__'s first argument expects text, the second argument is for images.) Please find the corrected code below:

import torch
from transformers import CLIPImageProcessor, CLIPModel, CLIPTokenizer
from PIL import Image

# Load the CLIP model
model_ID = "openai/clip-vit-base-patch32"
model = CLIPModel.from_pretrained(model_ID)

preprocess = CLIPImageProcessor.from_pretrained(model_ID)

# Define a function to load an image and preprocess it for CLIP
def load_and_preprocess_image(image_path):
    # Load the image from the specified path
    image = Image.open(image_path)

    # Apply the CLIP preprocessing to the image
    image = preprocess(image, return_tensors="pt")

    # Return the preprocessed image
    return image

# Load the two images and preprocess them for CLIP
image_a = load_and_preprocess_image('/content/bla.png')["pixel_values"]
image_b = load_and_preprocess_image('/content/bla.png')["pixel_values"]

# Calculate the embeddings for the images using the CLIP model
with torch.no_grad():
    embedding_a = model.get_image_features(image_a)
    embedding_b = model.get_image_features(image_b)

# Calculate the cosine similarity between the embeddings
similarity_score = torch.nn.functional.cosine_similarity(embedding_a, embedding_b)

# Print the similarity score
print('Similarity score:', similarity_score.item())

Output:

Similarity score: 1.0000001192092896

I found the several code in several places. looks like people are writing code without even running them and sharing. thank you so much it does work. can we say that this is the best method to calculate similarity at the moment? — Furkan Gözükara, Mar 11 '23 at 21:55
I generally hesitate to say that something is the best because there are many aspects you need to consider (e.g. performance, inference time ...). On the performance side, meta has published [slip](https://arxiv.org/abs/2112.12750) which is supposed to be better but it performed worse for the tasks I had to do. CLIP is still state-of-the-art, but hard to say if it is the best. @FurkanGözükara — cronoik, Mar 11 '23 at 22:18

How to calculate image similarity of given 2 images by using open AI Clip model - which method / AI model is best for calculating image similarity?

1 Answers1