Behind 145 camera feeds sits AI, scoring the CIIE kitchens in real time

As the ninth China International Import Expo (CIIE) is about to open, with the empowerment of new technologies led by artificial intelligence and the Internet of Things, regulation is more efficient and safety guarantees are stronger. Before last year's CIIE, a "Food Safety AI Doctor" service function independently developed by the Shanghai market regulation authorities had already been embedded in the CIIE market regulation service and support comprehensive command platform. A year later, this function has been fully upgraded. Today (November 2), in the on-site command center for CIIE market regulation service and support, a video screenshot caught the attention of a Jiefang Daily·Shangguang News reporter. This was a photo of the kitchen of a collective catering unit, randomly captured by an AI inspector from the live feed of the transparent kitchen monitoring system. Almost in sync with the capture time, the AI inspector gave the catering unit a "high risk" rating of below 80 points. The score was not given arbitrarily; it was the result obtained by the AI inspector comparing the on-site conditions one by one against relevant food safety norms and standards and deducting points for non-compliant items. The main problem with this unit lay in "process hygiene." In the screenshot, it could be clearly seen that a large amount of food debris was scattered on the floor of the kitchen work area, creating serious hygiene blind spots that could easily breed bacteria and lead to pest problems. In addition, some turnover baskets were placed directly on the ground, increasing the risk of cross-contamination. The scoring result was sent to the operator immediately in the form of a risk warning notice. Seeing the photos, the basis for the deductions, and specific rectification suggestions such as "immediately carry out deep cleaning of the floor" and "turnover boxes should be stored on shelves or pallets," the operator was fully convinced and immediately contacted the on-site person in charge to carry out rectification. The kitchen problem of a certain catering unit was captured. Photo by Chen Xihan. The on-site command center for CIIE market regulation service and support. Photo by Chen Xihan. The on-site command center for CIIE market regulation service and support. Photo by Chen Xihan. "In the past, transparent kitchen monitoring had a technical bottleneck, namely that problems in the footage had to be spotted manually. Most transparent kitchen systems only play and store footage and do not have the ability to automatically identify problems," admitted Qiu Congqian, deputy director of the office of the leading group for market regulation service and support at the ninth CIIE. Traditional manual regulation models have limits to their capabilities, while AI technology is in theory tireless. It can monitor catering units equipped with transparent kitchen systems 24 hours a day, and it can also multitask, simultaneously monitoring multiple feeds, identifying problems, issuing warnings, and making suggestions. With the help of new technologies such as artificial intelligence, regulators can be freed from a large amount of repetitive, inefficient "screen-watching" work and instead focus their energy on daily inspections and the high-risk targets identified during AI inspections. According to reports, AI inspectors have now "taken their seats" behind 145 monitoring cameras, conducting real-time regulation of 55 catering units and 4 collective catering enterprises within the CIIE venue. They keep a close eye on more than 30 kitchen operation details in three major areas: environmental hygiene, process hygiene, and personnel hygiene. They assign scores and then classify monitored targets into three risk levels: low, medium, and high. For high-risk units, regulators immediately visit to inspect and provide guidance; for medium-risk units, daily follow-up rechecks are conducted; for low-risk units, regulation is mainly conducted through off-site guidance, with "no disturbance unless necessary." During regulators' daily patrols, AI inspectors are also actively assisting. Among the CIIE market regulation service and support team, every staff member has a tool called "Snap and Verify" installed on their phone. Its principle is the same as that of the AI inspector in the transparent kitchen system: regulators take a photo on site, and the AI inspector automatically "picks out problems," helping regulators quickly and accurately pinpoint issues. "Manual on-site inspections are serial, meaning that after inspecting one target and making a judgment, one then inspects the next target. AI is parallel, viewing multiple targets at the same time," Qiu Congqian said. With AI's "initial screening," regulators can concentrate their energy on on-site problems that are "hidden deeper," such as opening refrigerators to check the interior, or checking corners and edges and blind spots of monitoring cameras, or moving closer to read the ingredients and shelf life on packaging bags. Qiu Congqian admitted that this AI system is still being "polished" to this day, and the current initial recognition accuracy rate has risen to 90.5%. "We also need to continuously feed photos of problems from standard regulatory enforcement procedures to the model so that AI can associate each photo with professional and abstract legal provisions and standards and norms, and know what is right and what is wrong." In the future, with the continuous upgrading of the AI system, even for complex issues such as whether a "reflective substance" on the floor is standing water, oil stains, or cleaning agent, AI inspectors will be able to immediately give more accurate judgments through the screen. "Snap and Verify." Photo by Chen Xihan.