This week’s list of data news highlights covers February 3 – 9, 2018, and includes articles about using the Apple Watch to detect diabetes and an AI system that can make sense of privacy policies.
Facebook researchers have developed an AI system called DensePose that can map a 2D image onto a human in video, allowing developers to overlay animations, character models, or other visual effects. The researchers generated over five million annotated data points based on 50,000 images of human body parts to train DensePose to identify human models in videos without the need for a depth sensor.
Researchers at the university of Pennsylvania have developed a brain stimulation technique that uses AI to boost human test subjects’ ability to memorize a series of words by 15 percent. The researchers recorded the brain activity of human subjects as they attempted to commit a series of words to memory. Then, using machine learning, the researchers were able to develop a model for each person that could predict how likely a subject was likely to remember a particular word based on their brain activity patterns. If the model predicted a subject had over a 50 percent chance of forgetting a word, it triggered stimulation through an electrode implanted in a portion of the brain linked to memory, substantially increasing the subject’s recall.
San Francisco autonomous vehicle startup Embark has successfully completed a test drive of a self-driving semi-truck travelling 2,400 miles from Los Angeles to Jacksonville, Florida without using a human driver on the highway. Embark embeds self-driving technology into regular semi-trucks, and uses machine learning to map its route in real-time based on sensor data, whereas some other autonomous driving systems only use sensor data to augment pre-mapped routes.
Researchers at health diagnostic company Cardiogram and UCSF have successfully completed a clinical study demonstrating that the Apple Watch can detect diabetes in wearers with 85 percent accuracy. The study analyzed heart sensor data from 14,000 Apple Watch wearers and was able to identify that 462 of them had diabetes. Determining if someone has diabetes normally involves analyzing blood glucose levels, but researchers have previously demonstrated that resting heart rate and heart rate variability could be a significant predictor of the disease.
Louisville, Kentucky has developed a pilot program to dispatch automated surveillance drones to investigate gunshots detected by acoustic sensors on the ground. The system relies on the gunshot-detecting connected sensor network developed by ShotSpotter, already in use in many U.S. cities. By automatically dispatching drones to a gunshot, law enforcement could more quickly identify potential suspects and protect officers from dangerous situations. Louisville has submitted the pilot to the U.S. Federal Aviation Administration for approval.
The PACUNAM Foundation, a Guatemalan nonprofit devoted to research and heritage preservation, has completed the first phase of its LIDAR initiative, which aims to map over 5,000 square miles of Guatemala’s lowlands, and discovered a large network of buried Mayan ruins by using aerial LIDAR mapping. The initiative used LIDAR to map 800 square miles of an area called the Maya Biosphere Reserve to create the largest-ever archaeological LIDAR dataset, revealing hidden ruins from over 60,000 houses, irrigation systems, roads, and other structures.
A UK-based startup called Optalysys has developed a processor that uses lasers to perform mathematical functions, known as optical computing, and has partnered with a group of genomics researchers to test using the processor for analyzing genetic data. Optical computing involves encoding data onto light beams, and by making these beams interfere with each other, it could enable complex calculations to execute in just one step, whereas traditional processors would take many steps to execute calculations. The technique has long been considered prohibitively complicated, but Optalysys says its processor can perform complex calculations called Fourier transformations 10 times faster than leading processors while using 75 percent less energy.
The United Kingdom’s Home Office has developed an identity verification system that uses handheld fingerprint scanners to help police officers quickly identify people they have detained if they fail to provide identification information. Normally when police detain someone who refuses to identify themselves, or if the police officer believes they are using fake information, an officer has to take them into custody to have their fingerprints taken, which can be time consuming. Now, police can use the Home Office’s system, which links a portable fingerprint scanner to a smartphone to quickly cross-check a person’s fingerprints against national law enforcement and immigration databases.
The U.S. National Institute for Standards and Technology (NIST) has published its draft Internet of Things-Enabled Smart City (IES-City) Framework to encourage widespread adoption of common vocabularies, standards, evaluation criteria, and other features to support smart city interoperability. The IES-City Framework includes examples of smart city use cases, detailed analysis of smart city architectures, and a methodology to help plan and measure their smart city development.
Image: The Pug Father.