A system utilizing artificial intelligence, machine learning and computer vision can identify whether people are wearing masks, which could help ensure that everyone entering a Mass Rapid Transport (MRT) station is wearing a mask during the COVID-19 pandemic.
A team from National Chiao Tung University that developed the system on Monday said that preliminary tests carried out at campus cafeteria and classrooms showed 95 percent accuracy.
The team was led by Cheng Wen-huang (鄭文皇) and Shuai Hong-han (帥宏翰), a professor and assistant professor in the electrical engineering department.
Photo copied by Hung Mei-hsiu, Taipei Times
For the tests, those wearing masks were marked with a green square and those without were marked with a red square, it said.
Key characteristics of facial features are targeted, allowing the system to identify individuals even without a full facial capture, including if their head was down as they used a cellphone or if their face was covered by an arm as they slept at a desk, the team said.
This approach is innovative, as previous programs were “taught” to identify individuals by being fed large portrait photographs, which raised the risk of misjudgements if facial features are obstructed, if the size of the face is too small or the angle is wrong, Shuai said.
In such circumstances, individuals would have to stand in front of a camera for a system to be able to recognize them, Shuai said.
The system could help monitor whether people are wearing masks on MRT trains or other forms of public transportation, as the government does not have enough people to monitor enforcement of the mask policy, the team said.
Using the system would not cause a back-up at entrances to MRT stations, as it retains high identification rates even with dense crowds, with results in under 0.03 seconds, the team said.
Meanwhile, National Cheng Kung University Hospital on Monday unveiled a lightweight protective tent designed to prevent the risk of infection when medical personnel are treating a COVID-19 patient.
Fang Pin-hui (方品惠), a physician with the Tainan-based hospital’s emergency medicine department who helped develop the tent, said doctors must be protected, especially when trying to intubate COVID-19 patients experiencing respiratory failure, which can expose them to aerosol-based transmission of the virus, especially if they are not wearing protective gear.
The tent-like device was inspired by an umbrella and a raincoat and consists of two L-shaped frames and a transparent PVC film, materials which are easy to find and relatively inexpensive, Fang said.
It can be placed on a bed or stretcher and acts as a barrier between a physician and patient, with holes cut in the film to allow doctors or nurses to put their hands through to perform procedures, she said.
Although intubation is typically not performed in an ambulance, the device could also be used by ambulance crews in an emergency to prevent them from being exposed to aerosols, Fang said.
The hospital said the tent was designed so that it would be easy to assemble, adding that the device was described in an article in the April issue of the Annals of Emergency Medicine.
Lai Hsien-yung (賴賢勇), an anesthesiologist at Mennonite Christian Hospital in Hualien, designed something similar in March — an acrylic box — which has already been adopted in the Philippines and Indonesia.
A small number of Taiwanese this year lost their citizenship rights after traveling in China and obtaining a one-time Chinese passport to cross the border into Russia, a source said today. The people signed up through Chinese travel agencies for tours of neighboring Russia with companies claiming they could obtain Russian visas and fast-track border clearance, the source said on condition of anonymity. The travelers were actually issued one-time-use Chinese passports, they said. Taiwanese are prohibited from holding a Chinese passport or household registration. If found to have a Chinese ID, they may lose their resident status under Article 9-1
Taiwanese were praised for their composure after a video filmed by Taiwanese tourists capturing the moment a magnitude 7.5 earthquake struck Japan’s Aomori Prefecture went viral on social media. The video shows a hotel room shaking violently amid Monday’s quake, with objects falling to the ground. Two Taiwanese began filming with their mobile phones, while two others held the sides of a TV to prevent it from falling. When the shaking stopped, the pair calmly took down the TV and laid it flat on a tatami mat, the video shows. The video also captured the group talking about the safety of their companions bathing
PROBLEMATIC APP: Citing more than 1,000 fraud cases, the government is taking the app down for a year, but opposition voices are calling it censorship Chinese Nationalist Party (KMT) Chairwoman Cheng Li-wun (鄭麗文) yesterday decried a government plan to suspend access to Chinese social media platform Xiaohongshu (小紅書) for one year as censorship, while the Presidential Office backed the plan. The Ministry of the Interior on Thursday cited security risks and accusations that the Instagram-like app, known as Rednote in English, had figured in more than 1,700 fraud cases since last year. The company, which has about 3 million users in Taiwan, has not yet responded to requests for comment. “Many people online are already asking ‘How to climb over the firewall to access Xiaohongshu,’” Cheng posted on
A classified Pentagon-produced, multiyear assessment — the Overmatch brief — highlighted unreported Chinese capabilities to destroy US military assets and identified US supply chain choke points, painting a disturbing picture of waning US military might, a New York Times editorial published on Monday said. US Secretary of Defense Pete Hegseth’s comments in November last year that “we lose every time” in Pentagon-conducted war games pitting the US against China further highlighted the uncertainty about the US’ capability to intervene in the event of a Chinese invasion of Taiwan. “It shows the Pentagon’s overreliance on expensive, vulnerable weapons as adversaries field cheap, technologically
RSS