These requirements are the nightmare of designers!
There are some common problems in design because of the high threshold of design and the different aesthetic standards. For example, when designing, it is found that the materials are not enough and there is no unified standard that leads to the dissatisfaction of customers with the design results, especially that some “advanced” designs can not resonate with consumers at all, which brings great challenges to the daily design work.
A large number of human cost, repeated operation cost of machinery, communication cost of repeated modification, cost of purchasing font and material copyright and other costs are consumed, which increases the burden of the enterprise, but also brings some infringement problems and risks.
Recently, JD cloud, together with JD retail user experience design department, launched professional intelligent design services to provide solutions to design efficiency, quality and consistency problems in the process of enterprise operation. Integrate JD’s own professional capabilities in design, big data, AI technology and other fields, build an intelligent design engine, help enterprises improve efficiency, save costs, optimize output, and provide capacity support for data delivery and operation.
The user will input the content that needs to be parsed, and the intelligent design engine will identify and analyze the input content, and then process it through the intelligent design processor to complete the operation required by the user, and finally produce the design result. Replace the designer’s manual design and operation, standardize, refine and automate the design results, and stabilize the output.
At present, intelligent design service provides services in the form of API interface on JD cloud official website. Let’s see what practical functions are available to help enterprises improve design efficiency and reduce design cost quickly!
The intelligent matting service is an intelligent service to remove background for image subject. Based on machine learning and image intelligent recognition technology, image processing efficiency can be greatly improved and labor cost can be reduced.
1. Features of intelligent matting function
Through the machine intelligence recognition image, does not need the artificial processing, only needs to process the picture upload, the extremely fast automatic output does not have the background matting result. Area division, zero threshold, low cost, no need to use design software, high matting success rate, continuous learning and Optimization Based on JD’s massive commodity image training machine, overall matting, especially for e-commerce commodity image matting success rate industry leading.
2. Intelligent matting application scenario
Whether it’s shooting by yourself or searching for materials from the Internet, it needs some processing before it can be used in image design. With the intelligent matting function, you can easily and quickly separate the main body of the picture from the background to generate a transparent bottom picture. At the same time, image optimization can be carried out for specific scenes, and the background needs to be changed according to the business situation, emphasis on key points and level differentiation. The image subject and background can be separated through intelligent matting service, and the processing of original materials can be completed quickly by the machine in batches.
The intelligent matting service currently serves the SKU Library of JD. It processes the commodity Library of JD in batches and builds a transparent bottom map material pool of commodities. Before that, it has accumulated 16 million pieces of commodity materials for various businesses of JD.
Intelligent color matching service is a practical function which is extended from a single color scheme to multiple color matching results.
Based on the image intelligent recognition technology, the image color can be accurately identified and replaced, with strong usability, which can quickly derive a variety of design styles, greatly expand the scope of use of pictures, and effectively reduce the cost of artificial production.
1. Advantages of intelligent color matching products
Intelligent color recognition and gray-scale calculation are carried out for the design object to quickly output color matching results of different styles in large quantities.
- Simple and easy to useMachine work instead of manpower, accurate, efficient and easy to use.
- Massive scheme: after flexible combination of materials, the number of color filling results is far beyond that of manual production.
- Strong usability: the color scheme of the system is preset in advance, which has been optimized for many times and the result is beautiful and reliable.
- Multi scene application: covering e-commerce, retail, logistics, media and other business scenarios.
- Support customization: customized development according to user requirements.
2. Application scenario of intelligent color matching
A picture can extend a wealth of color matching results. Users can choose the most appropriate color matching results according to the content of the picture. There is no need to test color manually, intuitively feel different color matching effects, assist decision-making, and select the best results.
The intelligent color matching function can perfectly divide the color filling area, support the custom color matching of pictures, and meet the needs of users to adjust the specific colors of pictures in batches, so as to unify the color matching of all pictures.
Style recognition service is able to read image information and realize the function of intelligent classification management according to style.
It is built by the deep learning engine. By building the sample data set training machine, it forms the style prediction model. According to the model, it can identify the style feature elements from the picture information and automatically identify the picture design style.
1. Features of style recognition function
Recognize and analyze any image transmitted by users, automatically divide the area, and intelligently recognize the image style feature elements according to the style prediction model. According to the feature elements, the pictures are classified according to the style, and the design style of the pictures is returned to the user.
The accuracy of recognition results is high. Based on the massive image data of JD, the accuracy of style prediction results has exceeded 83% when testing the advertising image data in e-commerce scenarios. The image recognition speed is fast, any style has no difference and accurate recognition, and the recognition result is returned in time, instead of manually viewing the style, judging the thinking process, eliminating the subjective judgment difference, and effectively unifying.
2. Application scenario of style recognition
The biggest application scenario is the intelligent management of pictures and the intelligent matching with users. It can be applied to the intelligent reading and tagging classification of pictures, and it is easy to achieve batch management without manual classification.