From 10fab319ebaa646c29c2c7ddb672acb29270ac69 Mon Sep 17 00:00:00 2001 From: PahudPlus <64403786+PahudPlus@users.noreply.github.com> Date: Tue, 7 Jul 2020 15:49:13 +0800 Subject: [PATCH] Removed outdated application.md (#18) --- site/en/menuStructure/en.json | 9 +-- site/en/reference/application.md | 114 --------------------------- site/zh-CN/menuStructure/cn.json | 9 +-- site/zh-CN/reference/application.md | 117 ---------------------------- 4 files changed, 2 insertions(+), 247 deletions(-) delete mode 100644 site/en/reference/application.md delete mode 100644 site/zh-CN/reference/application.md diff --git a/site/en/menuStructure/en.json b/site/en/menuStructure/en.json index 4814372c8..c95994d9c 100644 --- a/site/en/menuStructure/en.json +++ b/site/en/menuStructure/en.json @@ -206,20 +206,13 @@ "label3": "", "order": 5 }, - { - "id": "application.md", - "title": "Application Scenarios", - "label1": "reference", - "label2": "", - "order": 6 - }, { "id": "terms.md", "title": "Milvus Terminology", "label1": "reference", "label2": "", "label3": "", - "order": 7 + "order": 6 }, { "id": "faq", diff --git a/site/en/reference/application.md b/site/en/reference/application.md deleted file mode 100644 index 1e72a85ae..000000000 --- a/site/en/reference/application.md +++ /dev/null @@ -1,114 +0,0 @@ ---- -id: application.md -title: Application Scenarios -sidebar_label: Application Scenarios ---- - -# Application Scenarios - -## Typical scenarios - -Milvus can be used to build intelligent systems in most AI application scenarios: - -- Image search - - Query by image content，including content-based image retrieval such as bio-identification, object detection and recognition, etc. - -- Video processing - - Real-time object detection and tracing. - -- Natural language analysis - - Semantics-based text analysis and suggestion, and text similarity search. - -- Voiceprint recognition and audio search - -- Remove duplicated files by file fingerprint - -## Application architecture - -The application architecture of Milvus as a feature vector search engine is as follows: - -![MilvusApplication](../../../assets/application_arch.png) - -Unstructured data (images/videos/texts/audios) are transformed to feature vectors by feature extraction models, and saved to Milvus database. When you input a target vector, it is saved to the current vector collection, and the search begins, until the most similar vectors are matched, and their IDs returned. - -## Use case 1 - Personalized recommendation - -#### Background - -Nowadays, when you shop or view pages online, you will often see such words as "You may also like" or "Related products". In fact, many tech companies have embedded recommendation algorithms into their mobile Apps. Some examples include the Toutiao news, NetEase news, Pinduoduo, and WeChat, etc. With Milvus, you can implement your own personalized recommendation system. - -#### User requirements - -Recommend personalized content based on user persona. - -#### Application - -Take personalized advertising content recommendation as an example, the application architecture is: - -![Recommendation](../../../assets/ads_recommend.png) - -1. Create user persona by data analysis and key feature extraction - - By analyzing user history data and extracting key features, the user persona can be built. For example: The user history data contains news content about tennis, Wimbledon Championships, sports and Tennis Masters. So we can conclude from these keywords that the user is a tennis fan. - -2. Convert user keywords to vectors, load them to Milvus, and extract user feature vectors. - -3. Recommend content to users based on feature vectors and logistic regression model. - - 1) Search and filter out the top 100 ads that the user might be interested in and has not yet viewed. - 2) Extract the keywords and click-through rate of the top 100 ads. - 3) Locate and recommend the ads content to the user based on logistic regression model (which arises from user history data). - -## Use case 2 - Product feature extraction and multimodal search - -#### Background - -Online sellers need to prepare product images and tag product categories to help buyers better learn the product. As product categories grow, there will be a large sum of product images to be managed. If these product images are not well organized and utilized, it is often the case that you can't find the previously prepared image and need to retake it. - -#### User requirements - -Manage product images, and run multimodal similarity search based on keywords, for example, find out the most similar images of the most popular products. - -#### Application - -Milvus helps you realize product feature extraction and multimodal search by the following procedures: - -1. Convert product images to vectors. - -2. Load these vectors, together with other structured data such as product prices, publish date, sold quantity into Milvus. - -3. Begin multimodal search, specifying the query range as "among the top 10 products that sold the most". - -4. Find out the most similar images that belong to the top 10 products. - - -## Use case 3 - Video deduplication - -#### Background - -Today, online shopping and product trading has becoming a daily routine. On commodity trading platforms such as Taobao and Xianyu, sellers can display products to customers more fully and intuitively through product videos. Meanwhile, product video copying and plagiarism have also appeared. One solution to find a duplicate video is by vector similarity search. - -Take Xianyu, the second-hand commodity trading platform as an example. According to its current product size and business development trend, the vector index system needs to support billions of videos with an average length of 20 seconds, and a 1024-dimensional vector per second. - -#### User requirements - -Recognize and remove duplicate videos - -#### Application - -The core of video deduplication is high-dimensional vector index. Milvus helps you recognize duplicate videos through these steps: - -1. Video vectorization - - Convert video data to vectors according to certain algorithms. The converting algorithm determines how precisely the original video is represented by vectors. - -2. Vector distance computation - - When the video is represented by vectors, the similarity of videos can be measured by similarity of vectors. The distance between vectors can be calculated by the angle cosine, Euclidean distance and vector inner product. - -3. Vector index - - Search the most similar vector by multiple vector indexing methods such as tree-based, hash-based and vector quantization, etc. diff --git a/site/zh-CN/menuStructure/cn.json b/site/zh-CN/menuStructure/cn.json index 03d032296..7147ff570 100644 --- a/site/zh-CN/menuStructure/cn.json +++ b/site/zh-CN/menuStructure/cn.json @@ -206,20 +206,13 @@ "label3": "", "order": 5 }, - { - "id": "application.md", - "title": "应用场景", - "label1": "reference", - "label2": "", - "order": 6 - }, { "id": "terms.md", "title": "Milvus 术语", "label1": "reference", "label2": "", "label3": "", - "order": 7 + "order": 6 }, { "id": "faq", diff --git a/site/zh-CN/reference/application.md b/site/zh-CN/reference/application.md deleted file mode 100644 index 72abbd7ab..000000000 --- a/site/zh-CN/reference/application.md +++ /dev/null @@ -1,117 +0,0 @@ ---- -id: application.md -title: Application Scenarios -sidebar_label: Application Scenarios ---- - -# 应用场景 - -## 典型场景 - -在目前大部分的 AI 应用场景下，都可以使用 Milvus 来搭建智能应用系统： - -- 图片识别 - - 以图搜图，通过图片检索图片。具体应用例如：车辆检索和商品图片检索等。 - -- 视频处理 - - 针对视频信息的实时轨迹跟踪 - -- 自然语言处理 - - 基于语义的文本检索和推荐，通过文本检索近似文本。 - -- 声纹匹配，音频检索。 - -- 文件去重，通过文件指纹去除重复文件。 - - -## 典型架构 - -Milvus 做特征向量检索时典型应用架构如下： - -![MilvusApplication](../../../assets/application_arch_cn.png) - -非结构化数据（图像/视频/文字/音频等）首先通过特征提取模型产生特征向量，然后存入Milvus数据库系统。查询的时候，待查询的非结构化数据，也需要通过特征提取模型，提取特征向量。然后用该向量到Milvus中已存入的向量集里，查询匹配度最高的向量集合。最后，使用返回的向量ID，找到对应非结构化数据，结合上层应用，实现对应功能。 - -## 案例 1 - 个性化推荐系统 - -#### 背景 - -互联网时代个性化推荐已经渗透到人们生活的方方面面，例如常见的“猜你喜欢”、“相关商品”等。目前很多成功的手机 APP 都引入了个性化推荐算法，实时精准地把握用户兴趣，推荐他们最感兴趣的内容。例如，新闻类的有今日头条新闻客户端、网易新闻客户端等；商品广告类的有拼多多、微信等。Milvus 向量分析可以帮助你实现上述个性化推荐系统。 - -#### 用户需求 - -基于用户画像推荐个性化内容 - -#### 实现方案 - -以个性化广告内容推荐为例，Milvus 实现架构如下： - -![Recommendation](../../../assets/ads_recommend_cn.png) - -具体实现步骤为： - -1. 分析用户数据，找出关键词，构建相应用户画像。 - - 通过分析用户历史浏览数据，从中提取关键词，构建相应的用户画像。例如，某用户浏览了多条美容护肤产品或关注了相关公众号，浏览内容中包含了美容文章、护肤产品、防晒、美白等关键词，通过这些关键词可以得出该用户是一个热衷美容护肤的人。 - -2. 将用户关键词转换为向量，并将它们导入 Milvus，得到用户特征向量。 - -3. 基于用户特征向量，结合逻辑回归模型，将用户感兴趣的广告推荐给用户。 - - 1）Milvus 可以从互联网检索出前100条用户没有浏览过的广告，但是这100条广告却是该用户最感兴趣的广告。 - 2）从这100条广告中提取每条广告的关键词和点击率。 - 3）根据逻辑回归模型（该模型来自于用户以往的浏览的历史记录中），将用户感兴趣的广告推荐给用户。 - -## 案例 2 - 商品属性提取与多模搜索 - -#### 背景 - -为了方便买家更好地了解商品，电商卖家通常需要提供商品照片、标注商品类别和属性。随着商品种类的增加，将积累大量的图片素材。如果不能很好地管理这些图片，则容易出现找不到之前已经准备好的图片，需要重新拍摄的情况。 - -#### 用户需求 - -管理商品图片。根据关键词，对相似图片进行多模搜索。比如，搜索与目标图片最相似，且最近畅销度最高的所有商品图片。 - -#### 实现方案 - -Milvus 主要通过以下步骤实现商品属性提取与多模搜索： - -1. 将商品图片转化为向量。 - -2. 连同其它商品数据如价格、上市日期、卖出件数等结构化数据一并存入 Milvus。 - -3. 启动多模检索，并指定搜索范围为“卖出件数最多的商品”。 - -4. 在最畅销商品图片中中搜索出相似度最高的图片。 - - -## 案例 3 - 视频去重 - -#### 背景 - -如今，在线商品交易已经成为人们购物的日常，在诸如淘宝、咸鱼等商品交易平台上，卖家可以通过商品视频来更全面直观地向顾客展示商品。但与此同时也出现了一些视频拷贝、抄袭等不好的现象。其中一种解决方案时通过向量检索视频相似性，进而判断视频是否重复。 - -以二手商品交易平台闲鱼为例，根据其当前商品规模及业务发展的预估，闲鱼向量检索系统需支持检索亿级别平均时长为20秒，每秒向量维度是1024维的视频。 - -#### 用户需求 - -去除重复视频 - -#### 实现方案 - -视频去重本质是高维向量检索，Milvus 主要通过以下步骤实现： - -1. 视频向量化 - - 将视频数据按照一定的算法转换为向量，转换算法决定了向量表达原始视频数据的准确性。 - -2. 计算向量距离 - - 将视频转化为向量之后，计算视频的相似性就相当于计算向量的相似性。可以通过计算夹角余弦、欧氏距离和向量内积等方式计算向量间的距离。 - -3. 向量检索 - - 通过基于树的算法，哈希算法，矢量量化等，对向量进行检索，找出与目标向量（目标视频）相似度最高的向量。