Preface:
Search is a common way for people to get useful information from a huge amount of information.
Different from the daily “text search”, image search is to obtain the picture information people want through the way of “searching for pictures”.
At present, image search is rising rapidly with a wider range of applications and more vivid experience, gaining more and more market attention, and is recognized as one of the most potential applications in the current market, and has gradually been widely used in e-commerce, intellectual property protection, clothing, textiles, tourism and other fields.
Brief Introduction of speakers:
Ey, Alibaba data intelligent product operation experts, from Ali cloud intelligent business group.
The following content is compiled according to the video sharing and PPT of the speakers.
The main content is divided into three parts:
I. Market analysis: Image search is recognized as one of the applications with the most market potential. Two, product introduction: a picture wins thousands of words, from the interests and ideas directly to the search target. Third, user analysis: image search has been widely used in various industries.
I. Market analysis
From the perspective of the market, domestic e-commerce giants, such as Tmall and Taobao; There are more and more vertical e-commerce and cross-border e-commerce platforms, no matter the size of the platform, which are the potential target users of the image search product today.
Users’ demand for content retrieval when purchasing products is very strong and high, and it needs to be emphasized that: Their usage scenarios will not be in a quiet environment to buy, in commuting, at the time of the supermarket, there are you like a clothes, the results without the appropriate code number, this time you open taobao, a photo button search box, with photo, place the order directly, right on the second day the baby is at hand, you is it convenient for you to say ~ ~
This is our pealitao tool. We can directly search the corresponding pictures on Tmall and Taobao. If it is proved, the purchase conversion rate of Pealitao will be about 15% higher.
Another is the domestic gallery IP company. We have obtained a data from 2004 to 2016. We can directly see that the trading volume of picture materials has a trend of rapid growth. There are also many such clients in China, including Visual China, Oriental IC… There are also all kinds of pictures on the Internet, such as forum, or travel website those pictures also belong to the scope of our image search, these are also our potential customers!
two Image search product introduction
Image Search service (Image Search) is based on deep learning and machine vision technology, combining different industry applications and business scenarios.
Overview: to help users in the self-built image library to achieve the same or similar image search service.
Our products are based on the search service incubated by Alibaba e-commerce bealitao, providing low-cost, high-accuracy, easily delivered image search services! We are based on ali cloud IaaS bottom, and the algorithm maintenance cost is low, the use of learning cost is low, the use of monthly charges, at the same time unlimited number of calls, unlimited number of calls, unlimited number of calls!
Image search is based on tens of billions of level of image training, our algorithm is beyond doubt, especially in the e-commerce, retail field is invincible. Our service only considers the total size of customer pictures and QPS quantity, do not charge for the number of calls, simple and clear!
Product advantages: • High search accuracy: industry-leading search service. • High search efficiency: a combination of very large scale clustering and quantitative indexing technology to achieve millisecond response. • Support massive data: large-scale search engine can support tens of billions of levels of data. • Real-time interface: Supports real-time data addition and deletion. • Support customization: Customize development according to user requirements.
3. User analysis
Our target customers must have a large number of pictures, which can not only be photography works, but also various commodities. We can solve the core needs for customers is: product search, image search, similar image recommendation, etc.
As shown in the figure above, we can cover various categories, e-commerce category is our best. For other categories, we also have many mature cases, which can be fully commercialized. We should emphasize one point here: if there is a demand outside the above categories, such as cars, we can also do some exploration and customization.
Examples of graphic search scenarios:
E-commerce gallery scenario: Simplify the search process and improve the purchase experience. Users only need to take or upload photos of commodities to search for commodities according to the pictures. It eliminates the tedious text description, simplifies the commodity search process, and greatly improves the user’s shopping experience.
Image sharing and social networking websites usually have a large number of common images for users to search. By using image search service, an image search engine of 10 billion level can be quickly built on the cloud to provide the function of searching images by images and improve user experience.
One final word:
When we use graph search, there are two important processes:
1. Build an index library. The client will upload all the pictures offline, upload the photo library to the Ali Cloud for a feature extraction, after which we will build a search engine. 2. Online processes. When the new image comes, we need to do category recognition, subject detection, feature extraction, and then match with the original index built in the large library.
The above is the main content we want to tell you today, if you still do not understand the real demand for image search where, you can first think about the following three questions to try:
1. Can customers use accurate words to describe the product they are looking for? 2. How do customers distinguish between the details of desired products? 3. How do overseas customers describe requirements in different languages?
The original link
This article is the original content of the cloud habitat community, shall not be reproduced without permission.