Core demands
When the catering industry carries out commodity marketing, it usually faces difficulties in finding professional designers, high design costs, and complex production processes. Pu Meimei hopes to launch an intelligent design platform APP that can intelligently analyze users' intentions, simplify the design process to the greatest extent, and provide the best solution.
1. Intelligently analyze the user's demand for search terms. The intelligent analysis ability of conventional word segmentation technology is not enough, and it is difficult to correctly match the user's intention definition. I hope to seek more intelligent word segmentation technology.
2. Batch label the stock images on the platform to realize the search association between text and images. However, it is difficult to establish a label system due to the huge amount of pictures, heavy workload and low efficiency of manual labeling.
3. In addition to regular keyword search, there is a lack of more intelligent and convenient search methods, such as searching directly through images.
4. When selecting picture materials to make posters, if the existing pictures are directly rendered, it is easy to be ugly. You need to intelligently cut the picture materials before rendering.
Solution
1、 Lexical analysis to more accurately understand users' search intentions
Baidu lexical analysis provides three functions: word segmentation, part of speech tagging, and named entity recognition. It can identify the basic words (word segmentation) in the text string, reorganize these words, tag the part of speech of the combined words, and further identify the named entity to more accurately understand the user's search intention.
Step 1: Users input requirements through voice or text
Step 2: Call lexical analysis interface to resolve user semantics
Step 3: Carry out business processing according to the interface return result
2、 Universal image recognition, rapid construction of image label system
The platform is dedicated to providing poster design, menu production and other services for small and medium-sized offline catering businesses, with a huge amount of picture materials. Based on Baidu's general object and scene recognition services, we label images in batches, quickly build a background data label system, and further achieve accurate matching between text query and image meaning.
3、 Search pictures with pictures, a more intelligent and convenient search method
Step 1: The user clicks the Upload Picture button to select a picture;
Step 2: Call the similar image search interface to retrieve similar image results in the image library;
Step 3: Picture display and subsequent business operations are performed according to the interface return results.
4、 Image subject detection, automatic intelligent clipping
Detect the key subject position in the picture, intercept the subject position for rendering, and make the poster more beautiful.
Step 1: The user uploads a picture or selects a picture from the system;
Step 2: Call the image body detection interface to detect the main body content in the image (note: if it is a picture in the system, the main body position has been preset);
Step 3: Perform front-end rendering according to the main body location information returned by the interface, and present it to the user for subsequent business processing.
