This week’s list of data news highlights covers June 19, 2021 – June 25, 2021 and includes articles on restoring Rembrandt’s The Night Watch and identifying COVID-19 treatments.
Scientists at the Rijksmuseum, an art museum in Amsterdam, have completed restorations on Rembrandt’s painting The Night Watch using AI. In 1715, city officials damaged the painting when moving it through two doors in Amsterdam’s city hall. To recreate and restore those edges, scientists trained an AI system on a high-resolution scan of the original and a painted copy of the whole painting. The AI system completed the painting, adding three figures on the left side, a complete helmet on the right side, and a child in the left center.
Tour de France officials have partnered with technology company NTT to build a digital twin of the cycling race. Race officials will use sensors that give real-time updates about the race, providing a digital experience for fans, including insights on cyclist speed and position, and visibility for operations staff, enabling them to streamline operations across the 3,400 kilometer race and monitor adherence to COVID-19 guidelines.
Researchers at Johannes Kepler University Linz in Austria have developed a machine learning model that can detect missing people in dense forests. Rescue missions sometimes struggle to find people in forests because tree cover can obstruct their view and interfere with their heat sensors working accurately. The researchers’ system addresses this by using computer vision to see through trees and machine learning to determine whether the information from heat sensors comes from a human, animal, or other source.
Researchers at the University of Manchester, Wairakei Research Centre, and National Isotope Center in New Zealand have developed an algorithm to detect the early stages of a volcanic eruption using satellite data. The team input data from satellite images of volcanoes shortly before they erupted and ground data, such as gaseous emissions and tremors, into the algorithm to reconstruct the landscape before eruption. The algorithm found that volcanoes show observable signs of imminent eruptions up to 40 minutes ahead of time.
Researchers at ETH Zurich, a public university in Switzerland, have developed a machine learning model to measure the amount of carbon stored in underground vegetation. First, the researchers estimated the amount of global vegetation underground and then used machine learning to estimate how much carbon underground biomass stores based on existing data. The researchers found that underground vegetation contains 113 gigatonnes of carbon, or 10 years’ worth of global carbon dioxide emissions.
Researchers at the University of Tokyo have partnered with Japanese IT company Fujitsu to study COVID-19 treatments using Fugaku, the world’s most powerful supercomputer. Using Fugaku, researchers will analyze the properties of molecular compounds that could treat COVID-19, determine the impacts of different drugs on patients, and predict potential mutations of the virus.
Kheiron Medical Technologies, a medical technology company based in the United Kingdom, has developed an AI system that can analyze mammograms for signs of breast cancer as efficiently as a radiologist. The company trained the system on data from over three million mammography images and tested it on mammography screenings alongside human radiologists. The AI system and the human radiologist agreed on test results in 81.9 percent of cases.
The Boulder Police Department has launched an open data portal to provide information on crime in the city and guide its responses to service calls. The portal provides data on the number of service calls, the nature of the calls, and the neighborhoods from which the calls originate. For instance, the data shows that 366 bikes have been stolen this year, with the majority of bike thefts occurring in central Boulder.
Researchers at Sungkyunkwan University and Hanyang University in South Korea have created an adhesive film that uses AI to replicate the sense of touch. The film acts like skin by using sensors to collect data on the pressure, texture, and vibration of an object it touches. The sensors then transmit this information to an AI system to identify what object the film touched. In tests, the film achieved a texture classification accuracy rate of approximately 99 percent.
Walmart has developed an AI system to recommend substitutions to consumers for out-of-stock grocery orders. The AI system considers hundreds of variables, from the brand and price of a good to current inventory to individual customer preference. It presents substitutions to customers to approve or deny which, in turn, trains the AI system to become more accurate. Since its inauguration during the pandemic, the AI system has improved its customer acceptance rate from 90 percent to 97 percent.
Image credit: Flickr user Frans Vandewalle