(Original title: What else does AI recognize in addition to hotdogs? MIT announces the development of a new food identification technology) The Hot Dog Identification app “Not Hotdogâ€, previously developed by See Food Inc, sparked a lively discussion in the AI ​​world. However, the AI ​​technology involved in a seemingly simple and simplistic app is very complicated, although it is at the application level. Not much, but it is a milestone in the history of human AI. There have been reports of 36Kr before. Not Hotdog's reputation has triggered another round of AI food identification. MIT's Computer Science and Artificial Intelligence Laboratory recently studied this area. The MIT team intended to use the food identification system to identify the ingredients of the recipe from the food production video. The MIT team's pic2recipe (picture-to-receiver) system uses computer neural networks to determine the type of food in gourmet pictures on social networks, which can further analyze the uploader's health habits and dietary preferences. The Pic2recipe system uses the Food-101 Data Set, a food identification algorithm developed by Swiss scientists in 2014. It uses 101,000 food images in its database, which are cross-referenced to CSAIL's Recipe1M database data. Most of the data for the Recipe1M database was pulled from popular recipe websites such as All Recipes and Food.com. At present, the technology still has a long way to go from full maturity. The current system identification accuracy rate is only about 65%. The biggest bottleneck currently encountered by the project is the picture itself. Nick Hynes, a co-developer, said that when people take food photos, the presentation of food can be affected by the shooting status. Factors such as angle, distance, placement, and lighting may all cause differences in recognition results. When the same food appears in different recipes, the system's recognition error rate will also increase. The current system is better at identifying baked foods. Food has a trillion-dollar market, and it contains a large number of vertical fields. From retail to catering to social networking, etc., food can be seen. The kitchens, gourmets, and other food companies have all received investment in the food content website. It is believed that if the technology developed by MIT is mature, the imagination of the scenes that can be used is very large. Silicone Sheets,Silicone Gel Sheets,Thin Silicone Sheet,Silicone Sheet Nantong Boxin Electronic Technology Co., Ltd. , https://www.ntbosen.com
Zhang Yixi 