Dish and attribute information identification system and method based on fine-grained identification

Abstract

The invention discloses a dish and attribute information identification system and method based on fine-grained identification. The method comprises the following steps: preprocessing a dish image tobe identified; inputting the preprocessed image into a pre-trained deep neural network dish classification and attribute recognizer for recognition, and outputting the category attribute information of dishes; and matching the obtained category attribute information of the dishes with a pre-established dish database, and outputting an identification result. According to the invention, the relatedattribute information except for the varieties of vegetables is expanded, the existing identification accuracy is improved, and better systematic services are provided for the work of healthy diet recommendation, autonomous cooking, nutrition intake management and the like of people.

Description

technical field [0001] The invention belongs to the technical field of computer digital image processing and pattern recognition, and in particular relates to a vegetable type and attribute recognition system and method. Background technique [0002] With the improvement of people's living standards, it is very urgent for people to improve the material quality of life related to dishes, and a good dish identification method and system is the basis for realizing healthy meal recommendations, independent cooking, nutrient intake management, and self-service settlement in restaurants and cafeterias. Base. [0003] Fine-grained dish recognition is to identify the category and attribute information of hundreds of different kinds of dishes. The current fine-grained dish image recognition method is mainly based on the overall appearance information of the dish, extracting features through a convolutional neural network, and then performing classification and recognition tasks. Ex...

AI Technical Summary

Problems solved by technology

Excessive reliance on the types of dishes preset by the model, while ignoring the information of the ingredients of the dishes and the correspondence between the information of the ingredients and the dishes, has weak scalability for new dishes, and the robustness of dish recognition is poor. The results are not accurate enough, and the identification of attribute information is lacking, etc.

In addition, the current existing dish identification methods only consider the types of dishes, but do not fully consider the various attribute information of dishes.

Images