Vision Pack - Food Detection



Vision Pack Food Detection (VPF) service recognizes food items inside a refrigerator and converts the data into usable information.

This service uses image data taken by cameras installed in the DID (Door-in-Door) and the crisper drawer of a refrigerator as input values. It uses visual recognition technology to recognize food from the image data collected, and processes the data to provide the user with food names, types, and locations within the refrigerator. For example, if there is a can of cola in the DID, the service informs the user of the cola’s location. If the user puts an apple inside the crisper drawer, the service tells the user that an apple has been added to the crisper drawer.

VPF service can use its user learning function to register/edit/delete/identify a tag for the food that it recognized as a user's food. When the food is added to the refrigerator, the service recognizes it and assigns a relevant tag to the food. A user can edit the tag assigned to the food, and the service learns the food with the tag that the user-edited. Afterwards, when a food item similar to the learned food is added to the refrigerator, VPF service assigns the user's tag that it learned to the food. For example, when the user adds a “green juice” to the refrigerator, the service recognizes the food and assigns a “juice” tag. But, if the user changes the “juice” tag into a “diet juice” tag, the service learns the food with the tag. Afterwards, when the user adds a food item similar to the food that it learned to the refrigerator, the service assigns a “diet juice” tag, not a “juice” tag, to the food.


Process of VPF Engine


VPF service by LG AI Platform supports the following features:

Vision Pack 식품인식 엔진의 특징
Feature Description
Food recognition  

Recognizes more than 3,000 types of food:

  • Food inside the DID: 3,200+ types recognized
  • Produce inside the crisper drawer: 20+ types recognized
Various service applications

The service can be used for various services such as food management, barcode search, recipe recommendations, and online food purchases.

Ease of use

Since it supports Rest API, you can quickly develop food recognition functions using HTTP.

Excellent image recognition

The service uses the latest deep learning technology to provide the best recognition performance.

Fast response time

Fast response time with an optimized recognition service.

Wide recognition range

The service recognizes not only food items in the DID, but also produce in the crisper drawer.

User Learning

A user can register/edit/delete/identify a food tag.



VPF service receives image data as input values, extracts usable information from the input values, and sends the result to the application.


Architecture of VPF Engine


Examples of Use

VPF service can be used in various ways during everyday life.


  • Checks and manages the food in the refrigerator

The service checks the food in the refrigerator and informs the user of food with an approaching expiration date.

  • Recommends food recipes

The service recommends recipes for meals that utilize the food stored in the refrigerator.

  • Helps you order groceries online

If the food you need is not in the refrigerator, the service helps you place an online order with ease.