随笔- 228 文章- 1 评论- 15 阅读- 21万

如何实现图像搜索，文搜图，图搜图，CLIP+faiss向量数据库实现图像高效搜索

这是AIGC的时代，各种GPT大模型生成文本，还有多模态图文并茂大模型，

以及stable diffusion和stable video diffusion 图像生成视频生成等新模型，

层出不穷，如何生成一个图文并貌的文章，怎么在合适的段落加入图像，图像用什么方式获取，

图像可以使用搜索的形式获取，也可以使用stable diffusion生成

今天说说怎么使用搜索的形式获取，这种方式更高效，节省算力，更容易落地

clip模型，详细可以查看知乎

https://zhuanlan.zhihu.com/p/511460120

或论文https://arxiv.org/pdf/2103.00020.pdf

什么是faiss数据库

Faiss的全称是Facebook AI Similarity Search，是FaceBook的AI团队针对大规模相似度检索问题开发的一个工具，使用C++编写，有python接口，对10亿量级的索引可以做到毫秒级检索的性能。

简单来说，Faiss的工作，就是把我们自己的候选向量集封装成一个index数据库，它可以加速我们检索相似向量TopK的过程，其中有些索引还支持GPU构建，可谓是强上加强。

https://engineering.fb.com/2017/03/29/data-infrastructure/faiss-a-library-for-efficient-similarity-search/

1.huggingface下载clip模型，默认是英文版，也有中文版，英文版的效果会更好些

英文版

from PIL import Image
import requests
 
from transformers import CLIPProcessor, CLIPModel
 
model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
 
# url = "http://images.cocodataset.org/val2017/000000039769.jpg"
# image = Image.open(requests.get(url, stream=True).raw)
 
# inputs = processor(text=["a photo of a cat", "a photo of a dog"], images=image, return_tensors="pt", padding=True)
 
# image_features = model.get_image_features(inputs["pixel_values"])
# text_features = model.get_text_features(inputs["input_ids"],inputs["attention_mask"])
 
 
# outputs = model(**inputs)
# logits_per_image = outputs.logits_per_image # this is the image-text similarity score
# probs = logits_per_image.softmax(dim=1) # we can take the softmax to get the label probabilities
 
# print(probs)

中文版

from PIL import Image
import requests
from transformers import ChineseCLIPProcessor, ChineseCLIPModel
import torch
 
device = torch.device("mps")
 
model = ChineseCLIPModel.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")
processor = ChineseCLIPProcessor.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")
 
# url = "https://clip-cn-beijing.oss-cn-beijing.aliyuncs.com/pokemon.jpeg"
# image = Image.open(requests.get(url, stream=True).raw)
# Squirtle, Bulbasaur, Charmander, Pikachu in English
# texts = ["杰尼龟", "妙蛙种子", "小火龙", "皮卡丘"]
 
# # compute image feature
# inputs = processor(images=image, return_tensors="pt")
# image_features = model.get_image_features(**inputs)
# image_features = image_features / image_features.norm(p=2, dim=-1, keepdim=True)  # normalize
 
# # compute text features
# inputs = processor(text=texts, padding=True, return_tensors="pt")
# text_features = model.get_text_features(**inputs)
# text_features = text_features / text_features.norm(p=2, dim=-1, keepdim=True)  # normalize
 
# # compute image-text similarity scores
# inputs = processor(text=texts, images=image, return_tensors="pt", padding=True)
# outputs = model(**inputs)
# logits_per_image = outputs.logits_per_image  # this is the image-text similarity score
# probs = logits_per_image.softmax(dim=1)  # probs: [[1.2686e-03, 5.4499e-02, 6.7968e-04, 9.4355e-01]]

2.可以爬一些图片，做图像库，搜索也是在这个图像库中搜索，这个爬取的图像内容和业务场景相关，

比如你想获取动物的图像，那主要爬动物的就可以，这是我随便下载的一些图片

3.把图像映射成向量，存储在向量数据库faiss中

# from clip_model import model,processor
import faiss
from PIL import Image
import os
import json
from chinese_clip import model,processor
from tqdm import tqdm
 
d = 512
index = faiss.IndexFlatL2(d)   # 使用 L2 距离
 
# 文件夹路径
# folder_path = '/Users/smzdm/Downloads/Animals_with_Attributes2 2/JPEGImages'
folder_path = "image"
 
# 遍历文件夹
file_paths = []
for root, dirs, files in os.walk(folder_path):
    for file in files:
        # 检查文件是否为图片文件（这里简单地检查文件扩展名）
        if file.lower().endswith(('.png', '.jpg', '.jpeg', '.gif')):
            file_path = os.path.join(root, file)
            file_paths.append(file_path)
 
id2filename = {idx:x for idx,x in enumerate(file_paths)}
# 保存为 JSON 文件
with open('id2filename.json', 'w') as json_file:
    json.dump(id2filename, json_file)
 
for file_path in tqdm(file_paths,total=len(file_paths)):
    # 使用PIL打开图片
    image = Image.open(file_path)
    inputs = processor(images=image, return_tensors="pt", padding=True)
    image_features = model.get_image_features(inputs["pixel_values"])
    image_features = image_features / image_features.norm(p=2, dim=-1, keepdim=True)  # normalize
    image_features = image_features.detach().numpy()
    index.add(image_features)
    # 关闭图像，释放资源
    image.close()
 
faiss.write_index(index, "image.faiss")

4.加载数据库文件和索引文件，使用文本搜索图像或图像搜索图像

# from clip_model import model,processor
import faiss
from PIL import Image
import os
import json
from chinese_clip import model,processor
 
 
d = 512
index = faiss.IndexFlatL2(d)   # 使用 L2 距离
 
# 保存为 JSON 文件
with open('id2filename.json', 'r') as json_file:
    id2filename = json.load(json_file)
index = faiss.read_index("image.faiss")
 
 
def text_search(text,k=1):
    inputs = processor(text=text, images=None, return_tensors="pt", padding=True)
    text_features = model.get_text_features(inputs["input_ids"],inputs["attention_mask"])
    text_features = text_features / text_features.norm(p=2, dim=-1, keepdim=True)  # normalize
    text_features = text_features.detach().numpy()
    D, I = index.search(text_features, k)  # 实际的查询
 
    filenames = [[id2filename[str(j)] for j in i] for i in I]
 
    return text,D,filenames
 
def image_search(img_path,k=1):
    image = Image.open(img_path)
    inputs = processor(images=image, return_tensors="pt")
    image_features = model.get_image_features(**inputs)
    image_features = image_features / image_features.norm(p=2, dim=-1, keepdim=True)  # normalize
 
    image_features = image_features.detach().numpy()
    D, I = index.search(image_features, k)  # 实际的查询
 
    filenames = [[id2filename[str(j)] for j in i] for i in I]
 
    return img_path,D,filenames
 
 
 
if __name__ == "__main__":
 
    text = ["雪山","熊猫","长城","苹果"]
    text,D,filenames = text_search(text)
    print(text,D,filenames)
 
    # img_path = "image/apple2.jpeg"
    # img_path,D,filenames = image_search(img_path,k=2)
    # print(img_path,D,filenames)

比如用文字搜索

1	`["雪山","熊猫","长城","苹果"]返回结果：`

['雪山', '熊猫', '长城', '苹果'] [[1.2182312]
[1.1529984]
[1.1177421]
[1.1656866]] [['image/OIP (10).jpeg'], ['image/OIP.jpeg'], ['image/OIP (8).jpeg'], ['image/apple2.jpeg']]