ICCV 2021 | Meituan "LargeFineFoodAI" seminar is about to open, interactive awards

The long-awaited ICCV 2021 is here as promised. Adhering to the corporate mission of "Helping everyone eat better and live better", Meituan will focus on the application of computer vision technology in the field of large-scale and fine-grained food analysis.

From 19:00-22:30, October 16th, Beijing time, ’s Visual Intelligence Department will join hands with the Institute of Computing Technology of the Chinese Academy of Sciences, Beijing Zhiyuan and the University of Barcelona . "The theme of the exchange discussion.

The agenda of the seminar will be divided into the following three parts. Due to the impact of the epidemic, the whole process will be conducted online ( for details of participation methods, see at the bottom of the article).

01 Invited Talk

As one of the three top conferences in the direction of computer vision, the ICCV International Computer Vision Conference has focused on the fine-grained recognition and retrieval of large-scale food images with "LargeFineFoodAI" as the theme of the seminar. In the highly anticipated sharing session, we invited three top experts in the industry to bring the latest theories and practical results on the application of computer vision in the food field.

Food is an important source of physical health and spiritual happiness for everyone, but personal dietary preferences may not match their physical fitness. For most people, it is very difficult to strike a balance between healthy eating and happy eating. Ideally, you should build your own food model for everyone to customize their diet. Therefore, this sharing will focus on how to build a personal food model and food map.

A new type of food log tool, FoodLog Athl, was born, which can be used for diet-related health care and diet assessment services. From the perspective of nutritionists or monitoring users, this tool can support multiple functions such as food image recognition, nutritional diet evaluation, and food nutritional value calculation. In addition, this sharing will also introduce key scientific research results such as the role of the tool before and after COVID-19 and changes in related food statistics.

Neural networks have become one of the most powerful forecasting systems. Among them, the Bayesian paradigm of deep learning takes probability as the learning target of neural network architecture and parameters, and quantifies the uncertainty of prediction through the posterior distribution. In this sharing, we will discuss why uncertainty estimation is needed, and how to model and measure it. At the same time, in order to further demonstrate its application value, we will discuss how to use uncertainty for modeling in food identification.

02 Challenge Report

The challenge organized by this seminar with the theme of "large-scale food image recognition and retrieval" also attracted the participation of many powerful teams at home and abroad, including Tsinghua University, University of Science and Technology of China, Nanjing University of Science and Technology, University of Barcelona, Singapore Nanyang University of Science and Technology; 143 domestic and foreign teams including Alibaba, Shenlan Technology, OPPO, Huanju Times and other companies participated in the competition.

The competition is divided into two major courses: large-scale food image fine-grained recognition and large-scale food image fine-grained retrieval, and selection is made based on the final results and submitted technical solutions. The following winning teams in the seminar will also report the results of the competition and technical solutions to everyone.

Table 1. The ranking list of the top-3 teams in the “Large-scale fine-grained food recognition” challenge at ICCV 2021

Table 2. The ranking list of the top-3 teams in the “Large-scale fine-grained food retrieval” challenge at ICCV 2021

Benefits: The data set continues to be open

In this competition, we proposed a dataset containing more than 1,000 fine-grained food categories and more than 500,000 images, including Chinese and Western food. With the help of food experts, a unified food ontology was constructed by combining and adapting the existing food classification system. The number of images in each category is within the range of [153; 1999]. Compared with the existing food data set, it reflects a larger category imbalance, and it also brings greater challenges to recognition and retrieval! Taking advantage of the opportunity of this seminar and competition, we will continue to disclose the data set to provide valuable assistance to promote the application of computer vision in the field of food analysis.

Award-winning interaction

In the report session of the challenge on the 16th, online audiences participating in the questioning interaction will have the opportunity to win the following prizes. Looking forward to your questions and challenges!

03 Oral Presentation

According to official ICCV news, a total of 6,236 submissions were received this year, an increase of about 50% over the previous year; finally 1,617 papers were accepted, with an acceptance rate of 25.9%. The LargeFineFoodAI seminar called for food analysis on the subject, and received high-quality paper feedback. Finally, two papers from Carnegie Mellon University and Purdue University were accepted. The authors of the two papers will also be invited to bring wonderful reports and interpretations at the meeting.

People take food as their heaven. When we use computer vision technology to rediscover food, how can we make everyone eat better, more scientifically, and healthier? Looking forward to working with us to find the answer!

To participate in the conference, please press and hold or scan the QR code above to reply "LargeFineFoodAI", and you will be automatically added to the LargeFineFoodAI2021 technical exchange group.

Full agenda : https://foodai-workshop.meituan.com/foodai2021.html#index