RAG multimodal dengan Milvus

Tutorial ini menampilkan RAG multimodal yang didukung oleh Milvus, model BGE yang divisualisasikan, dan GPT-4o. Dengan sistem ini, pengguna dapat mengunggah gambar dan mengedit instruksi teks, yang diproses oleh model pengambilan yang disusun oleh BGE untuk mencari gambar kandidat. GPT-4o kemudian bertindak sebagai pemeringkat, memilih gambar yang paling sesuai dan memberikan alasan di balik pilihan tersebut. Kombinasi yang kuat ini memungkinkan pengalaman pencarian gambar yang mulus dan intuitif, memanfaatkan Milvus untuk pengambilan yang efisien, model BGE untuk pemrosesan dan pencocokan gambar yang tepat, dan GPT-4o untuk pemeringkatan ulang tingkat lanjut.

Persiapan

Instal Ketergantungan

$ 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

Unduh Data

Perintah berikut ini akan mengunduh data contoh dan mengekstrak ke folder lokal "./images_folder", termasuk:

• images: Bagian dari Amazon Reviews 2023 yang berisi sekitar 900 gambar dari kategori "Alat", "Ponsel dan Aksesori", dan "Elektronik".

• macantutul.jpg: Contoh gambar kueri.

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

Model Penyematan Muatan

Kami akan menggunakan model BGE yang divisualisasikan "bge-visualized-base-en-v1.5" untuk menghasilkan sematan baik untuk gambar maupun teks.

1. Berat unduhan

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

2. Bangun pembuat enkode

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)

Muat Data

Bagian ini akan memuat gambar contoh ke dalam database dengan embedding yang sesuai.

Menghasilkan embeddings

Muat semua gambar jpeg dari direktori data dan terapkan penyandi untuk mengonversi gambar menjadi embedding.

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

Masukkan ke dalam Milvus

Sisipkan gambar dengan jalur dan sematan yang sesuai ke dalam koleksi 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}

Pencarian Multimodal dengan Perangking Generatif

Pada bagian ini, pertama-tama kita akan mencari gambar yang relevan dengan kueri multimodal dan kemudian menggunakan layanan LLM untuk memberi peringkat ulang hasil dan menemukan yang terbaik dengan penjelasan.

Jalankan pencarian

Sekarang kita siap untuk melakukan pencarian gambar lanjutan dengan data kueri yang terdiri dari gambar dan instruksi teks.

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']

Beri peringkat ulang dengan GPT-4o

Kami akan menggunakan LLM untuk mengurutkan gambar dan menghasilkan penjelasan untuk hasil terbaik berdasarkan kueri pengguna dan hasil yang diambil.

1. Buat tampilan panorama

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

Gabungkan gambar kueri dan gambar yang diambil dengan indeks dalam tampilan panorama.

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()

Membuat tampilan panorama

2. Beri peringkat ulang dan jelaskan

Kami akan mengirimkan gambar gabungan ke layanan LLM multimodal bersama dengan petunjuk yang tepat untuk memberi peringkat hasil yang diambil dengan penjelasan. Untuk mengaktifkan GPT-4o sebagai LLM, Anda harus menyiapkan OpenAI API Key.

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

Dapatkan indeks gambar setelah pemeringkatan dan alasan hasil terbaik:

ranked_indices, explanation = generate_ranking_explanation(
    combined_image_path, query_text
)

3. Tampilkan hasil terbaik dengan penjelasan

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.

Hasil terbaik

Penerapan Cepat

Untuk mempelajari tentang cara memulai demo online dengan tutorial ini, silakan lihat contoh aplikasi.