RAG multimodal com Milvus

Este tutorial apresenta o RAG multimodal com o Milvus, o modelo BGE visualizado e o GPT-4o. Com este sistema, os utilizadores podem carregar uma imagem e editar instruções de texto, que são processadas pelo modelo de recuperação composto da BGE para procurar imagens candidatas. O GPT-4o actua então como um reranker, selecionando a imagem mais adequada e fornecendo a lógica por detrás da escolha. Esta combinação poderosa permite uma experiência de pesquisa de imagens intuitiva e sem falhas, tirando partido do Milvus para uma recuperação eficiente, do modelo BGE para um processamento e correspondência precisos das imagens e do GPT-4o para uma classificação avançada.

Preparação

Instalar dependências

$ pip install --upgrade pymilvus openai datasets opencv-python timm einops ftfy peft tqdm

$ git clone https://github.com/FlagOpen/FlagEmbedding.git
$ pip install -e FlagEmbedding

Descarregar dados

O comando a seguir descarregará os dados do exemplo e os extrairá para uma pasta local "./images_folder", incluindo:

• imagens: Um subconjunto de Amazon Reviews 2023 contendo aproximadamente 900 imagens das categorias "Appliance", "Cell_Phones_and_Accessories" e "Electronics".

• leopard.jpg: Um exemplo de imagem de consulta.

$ wget https://github.com/milvus-io/bootcamp/releases/download/data/amazon_reviews_2023_subset.tar.gz
$ tar -xzf amazon_reviews_2023_subset.tar.gz

Modelo de incorporação de carga

Vamos utilizar o modelo Visualized BGE "bge-visualized-base-en-v1.5" para gerar embeddings para imagens e texto.

1. Descarregar peso

$ wget https://huggingface.co/BAAI/bge-visualized/resolve/main/Visualized_base_en_v1.5.pth

2. Construir codificador

import torch
from FlagEmbedding.visual.modeling import Visualized_BGE


class Encoder:
    def __init__(self, model_name: str, model_path: str):
        self.model = Visualized_BGE(model_name_bge=model_name, model_weight=model_path)
        self.model.eval()

    def encode_query(self, image_path: str, text: str) -> list[float]:
        with torch.no_grad():
            query_emb = self.model.encode(image=image_path, text=text)
        return query_emb.tolist()[0]

    def encode_image(self, image_path: str) -> list[float]:
        with torch.no_grad():
            query_emb = self.model.encode(image=image_path)
        return query_emb.tolist()[0]


model_name = "BAAI/bge-base-en-v1.5"
model_path = "./Visualized_base_en_v1.5.pth"  # Change to your own value if using a different model path
encoder = Encoder(model_name, model_path)

Carregar dados

Esta secção irá carregar imagens de exemplo para a base de dados com os respectivos embeddings.

Gerar embeddings

Carregue todas as imagens jpeg do diretório de dados e aplique o codificador para converter as imagens em embeddings.

import os
from tqdm import tqdm
from glob import glob


# Generate embeddings for the image dataset
data_dir = (
    "./images_folder"  # Change to your own value if using a different data directory
)
image_list = glob(
    os.path.join(data_dir, "images", "*.jpg")
)  # We will only use images ending with ".jpg"
image_dict = {}
for image_path in tqdm(image_list, desc="Generating image embeddings: "):
    try:
        image_dict[image_path] = encoder.encode_image(image_path)
    except Exception as e:
        print(f"Failed to generate embedding for {image_path}. Skipped.")
        continue
print("Number of encoded images:", len(image_dict))

Generating image embeddings: 100%|██████████| 900/900 [00:20<00:00, 44.08it/s]

Number of encoded images: 900

Inserir no Milvus

Insere as imagens com os caminhos e os embeddings correspondentes na coleção Milvus.

from pymilvus import MilvusClient


dim = len(list(image_dict.values())[0])
collection_name = "multimodal_rag_demo"

# Connect to Milvus client given URI
milvus_client = MilvusClient(uri="./milvus_demo.db")

# Create Milvus Collection
# By default, vector field name is "vector"
milvus_client.create_collection(
    collection_name=collection_name,
    auto_id=True,
    dimension=dim,
    enable_dynamic_field=True,
)

# Insert data into collection
milvus_client.insert(
    collection_name=collection_name,
    data=[{"image_path": k, "vector": v} for k, v in image_dict.items()],
)

{'insert_count': 900,
 'ids': [451537887696781312, 451537887696781313, ..., 451537887696782211],
 'cost': 0}

Pesquisa multimodal com o Generative Reranker

Nesta secção, começamos por procurar imagens relevantes através de uma consulta multimodal e, em seguida, utilizamos o serviço LLM para classificar os resultados e encontrar o melhor com uma explicação.

Executar pesquisa

Agora estamos prontos para efetuar a pesquisa avançada de imagens com dados de consulta compostos por instruções de imagem e de texto.

query_image = os.path.join(
    data_dir, "leopard.jpg"
)  # Change to your own query image path
query_text = "phone case with this image theme"

# Generate query embedding given image and text instructions
query_vec = encoder.encode_query(image_path=query_image, text=query_text)

search_results = milvus_client.search(
    collection_name=collection_name,
    data=[query_vec],
    output_fields=["image_path"],
    limit=9,  # Max number of search results to return
    search_params={"metric_type": "COSINE", "params": {}},  # Search parameters
)[0]

retrieved_images = [hit.get("entity").get("image_path") for hit in search_results]
print(retrieved_images)

['./images_folder/images/518Gj1WQ-RL._AC_.jpg', './images_folder/images/41n00AOfWhL._AC_.jpg', './images_folder/images/51Wqge9HySL._AC_.jpg', './images_folder/images/51R2SZiywnL._AC_.jpg', './images_folder/images/516PebbMAcL._AC_.jpg', './images_folder/images/51RrgfYKUfL._AC_.jpg', './images_folder/images/515DzQVKKwL._AC_.jpg', './images_folder/images/51BsgVw6RhL._AC_.jpg', './images_folder/images/51INtcXu9FL._AC_.jpg']

Classificar novamente com GPT-4o

Vamos utilizar um LLM para classificar as imagens e gerar uma explicação para o melhor resultado com base na consulta do utilizador e nos resultados obtidos.

1. Criar uma vista panorâmica

import numpy as np
import cv2

img_height = 300
img_width = 300
row_count = 3


def create_panoramic_view(query_image_path: str, retrieved_images: list) -> np.ndarray:
    """
    creates a 5x5 panoramic view image from a list of images

    args:
        images: list of images to be combined

    returns:
        np.ndarray: the panoramic view image
    """
    panoramic_width = img_width * row_count
    panoramic_height = img_height * row_count
    panoramic_image = np.full(
        (panoramic_height, panoramic_width, 3), 255, dtype=np.uint8
    )

    # create and resize the query image with a blue border
    query_image_null = np.full((panoramic_height, img_width, 3), 255, dtype=np.uint8)
    query_image = Image.open(query_image_path).convert("RGB")
    query_array = np.array(query_image)[:, :, ::-1]
    resized_image = cv2.resize(query_array, (img_width, img_height))

    border_size = 10
    blue = (255, 0, 0)  # blue color in BGR
    bordered_query_image = cv2.copyMakeBorder(
        resized_image,
        border_size,
        border_size,
        border_size,
        border_size,
        cv2.BORDER_CONSTANT,
        value=blue,
    )

    query_image_null[img_height * 2 : img_height * 3, 0:img_width] = cv2.resize(
        bordered_query_image, (img_width, img_height)
    )

    # add text "query" below the query image
    text = "query"
    font_scale = 1
    font_thickness = 2
    text_org = (10, img_height * 3 + 30)
    cv2.putText(
        query_image_null,
        text,
        text_org,
        cv2.FONT_HERSHEY_SIMPLEX,
        font_scale,
        blue,
        font_thickness,
        cv2.LINE_AA,
    )

    # combine the rest of the images into the panoramic view
    retrieved_imgs = [
        np.array(Image.open(img).convert("RGB"))[:, :, ::-1] for img in retrieved_images
    ]
    for i, image in enumerate(retrieved_imgs):
        image = cv2.resize(image, (img_width - 4, img_height - 4))
        row = i // row_count
        col = i % row_count
        start_row = row * img_height
        start_col = col * img_width

        border_size = 2
        bordered_image = cv2.copyMakeBorder(
            image,
            border_size,
            border_size,
            border_size,
            border_size,
            cv2.BORDER_CONSTANT,
            value=(0, 0, 0),
        )
        panoramic_image[
            start_row : start_row + img_height, start_col : start_col + img_width
        ] = bordered_image

        # add red index numbers to each image
        text = str(i)
        org = (start_col + 50, start_row + 30)
        (font_width, font_height), baseline = cv2.getTextSize(
            text, cv2.FONT_HERSHEY_SIMPLEX, 1, 2
        )

        top_left = (org[0] - 48, start_row + 2)
        bottom_right = (org[0] - 48 + font_width + 5, org[1] + baseline + 5)

        cv2.rectangle(
            panoramic_image, top_left, bottom_right, (255, 255, 255), cv2.FILLED
        )
        cv2.putText(
            panoramic_image,
            text,
            (start_col + 10, start_row + 30),
            cv2.FONT_HERSHEY_SIMPLEX,
            1,
            (0, 0, 255),
            2,
            cv2.LINE_AA,
        )

    # combine the query image with the panoramic view
    panoramic_image = np.hstack([query_image_null, panoramic_image])
    return panoramic_image

Combinar a imagem de consulta e as imagens recuperadas com índices numa vista panorâmica.

from PIL import Image

combined_image_path = os.path.join(data_dir, "combined_image.jpg")
panoramic_image = create_panoramic_view(query_image, retrieved_images)
cv2.imwrite(combined_image_path, panoramic_image)

combined_image = Image.open(combined_image_path)
show_combined_image = combined_image.resize((300, 300))
show_combined_image.show()

Criar uma vista panorâmica

2. Classificar novamente e explicar

Enviaremos a imagem combinada para o serviço LLM multimodal, juntamente com os avisos adequados para classificar os resultados obtidos com uma explicação. Para ativar o GPT-4o como LLM, é necessário preparar a chave da API OpenAI.

import requests
import base64

openai_api_key = "sk-***"  # Change to your OpenAI API Key


def generate_ranking_explanation(
    combined_image_path: str, caption: str, infos: dict = None
) -> tuple[list[int], str]:
    with open(combined_image_path, "rb") as image_file:
        base64_image = base64.b64encode(image_file.read()).decode("utf-8")

    information = (
        "You are responsible for ranking results for a Composed Image Retrieval. "
        "The user retrieves an image with an 'instruction' indicating their retrieval intent. "
        "For example, if the user queries a red car with the instruction 'change this car to blue,' a similar type of car in blue would be ranked higher in the results. "
        "Now you would receive instruction and query image with blue border. Every item has its red index number in its top left. Do not misunderstand it. "
        f"User instruction: {caption} \n\n"
    )

    # add additional information for each image
    if infos:
        for i, info in enumerate(infos["product"]):
            information += f"{i}. {info}\n"

    information += (
        "Provide a new ranked list of indices from most suitable to least suitable, followed by an explanation for the top 1 most suitable item only. "
        "The format of the response has to be 'Ranked list: []' with the indices in brackets as integers, followed by 'Reasons:' plus the explanation why this most fit user's query intent."
    )

    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {openai_api_key}",
    }

    payload = {
        "model": "gpt-4o",
        "messages": [
            {
                "role": "user",
                "content": [
                    {"type": "text", "text": information},
                    {
                        "type": "image_url",
                        "image_url": {"url": f"data:image/jpeg;base64,{base64_image}"},
                    },
                ],
            }
        ],
        "max_tokens": 300,
    }

    response = requests.post(
        "https://api.openai.com/v1/chat/completions", headers=headers, json=payload
    )
    result = response.json()["choices"][0]["message"]["content"]

    # parse the ranked indices from the response
    start_idx = result.find("[")
    end_idx = result.find("]")
    ranked_indices_str = result[start_idx + 1 : end_idx].split(",")
    ranked_indices = [int(index.strip()) for index in ranked_indices_str]

    # extract explanation
    explanation = result[end_idx + 1 :].strip()

    return ranked_indices, explanation

Obter os índices de imagem após a classificação e o motivo do melhor resultado:

ranked_indices, explanation = generate_ranking_explanation(
    combined_image_path, query_text
)

3. Apresentar o melhor resultado com uma explicação

print(explanation)

best_index = ranked_indices[0]
best_img = Image.open(retrieved_images[best_index])
best_img = best_img.resize((150, 150))
best_img.show()

Reasons: The most suitable item for the user's query intent is index 6 because the instruction specifies a phone case with the theme of the image, which is a leopard. The phone case with index 6 has a thematic design resembling the leopard pattern, making it the closest match to the user's request for a phone case with the image theme.

O melhor resultado

Implementação rápida

Para saber como iniciar uma demonstração online com este tutorial, consulte a aplicação de exemplo.