Good news for healthy people, you can get food calories and nutrients by taking pictures

Nowadays, body management has become the focus of people's daily life. In addition to exercise, calorie intake is also the top priority. To maintain ideal physical health and weight status, people need to measure their daily intake for a long time. The volume, calorie and nutritional value of food, which requires practitioners to have strong patience, execution and knowledge reserves, have become a stumbling block on the road of body management.

Therefore, many sports and health apps support food recognition, which can realize the function of taking pictures to identify food calories and nutritional elements, meeting people's needs to quickly obtain food information and timely management of calorie intake in daily life, and provide great convenience for people's body management. .

Technical principle

So, what is the technical rationale behind such a function of taking pictures to obtain food calorie and nutritional information? In fact, it depends on the image classification ability .

Image classification ability is an important basic function in the field of AI artificial intelligence. It has a wide range of practical application scenarios. The traditional image classification method process includes preprocessing, feature extraction, and classifier. Among them, feature extraction requires researchers to invest a lot of effort in manual design and extraction, and only applies to For simple image classification, it is impossible to recognize the actual complex picture content.

In recent years, the ability of image classification based on deep learning has become popular. Using a specific reasoning framework, through the core technology of neural network, it can classify the entity objects in the image and add annotation information to help define the subject and applicable scene of the image.

The general picture classification process is as follows: the developer passes in the photo to use static image detection, or activates the camera to use video stream detection. The image classification capability will be analyzed according to the device-side/cloud-side algorithm model used by the developer, and returned to the App image. Category (eg: plants, furniture, mobile phones) and confidence.

Performance Advantages of HMS Core Machine Learning Service Image Classification Capability

The machine learning service image classification function of Huawei HMS Core is based on the deep learning method, which can identify objects, scenes, behaviors and other information in pictures, and return the corresponding label information. The machine learning service image classification capabilities have improved in recognition accuracy and speed performance.

Transfer learning method: Enhance the image labeling model and knowledge transfer ability, and continuously optimize the topology of the deep neural network, increasing the accuracy rate by 38%;

Integrated WordNet semantic network: optimize the semantic analysis model, perform semantic analysis on image content, realize automatic inference of concept labels, and support 23,000 labels;

Based on Huawei GPU cloud service acceleration: compared with the previous generation GPU, the memory bandwidth is increased by 2 times, the bit width is increased by 8 times, and it only takes 100 milliseconds to recognize a single image.

Developers can use the existing categories of image classification capabilities, or customize the image classification model, collect and train various food images, and then import the corresponding label data to form a huge food calorie database. Finally, through the depth-of-field camera function of the Huawei camera, the distance between the mobile phone and the object can be effectively measured, the size and weight of the object can be roughly identified, and then matched with the information in the database, so as to realize the calorie detection of different categories and sizes of food. calculate.

In addition, the machine learning service image classification capability is also widely used in image classification management scenarios, such as mobile phone album classification management, e-commerce App product map intelligent classification and other scenarios, providing good solutions for various application developers. .

Click to view a list of image recognition classification information .

Learn more details>>

Visit the official website of Huawei Developer Alliance
Get development guidance documents
Huawei Mobile Services Open Source Warehouse Address: GitHub , Gitee

Follow us to know the latest technical information of HMS Core for the first time~