AI researchers from more than 30 countries around the world came together this week for the AI City Challenge, a competition to spur the development of better machine learning via tasks such as detecting car accidents and tracking a vehicle across a network of cameras. Now in its fourth year, the challenge pushes AI researchers to create more efficient Intelligent Transportation Systems (ITS).
Teams from Baidu won three of the four competitions: vehicle counting, multi-camera reidentification, and car accident and stalled vehicle detection. Organizing committee member and University of Albany assistant professor Ming-Ching Chang said during the virtual workshop that the top performing model in this category achieved 95.3% accuracy.
A team from Carnegie Mellon University won one of the four challenges for tracking a vehicle over a network of multiple cameras. The benchmark data set for this challenge stretches across 46 camera views spanning 16 intersections in Dubuque, Iowa.
In total, the competition drew more than 800 individual researchers on 300 teams from 36 nations; 76 teams submitted code for final review. The organizing committee included companies like Amazon and Nvidia as well as researchers associated with Iowa State University, Santa Clara University, and the Indian Institute of Technology Kanpur. Organizers called this year’s AI City Challenge the first to use effectiveness and computation efficiency standards the U.S. Department of Transportation says it needs to consider deployment of this form of automation in the wild. Past competitions focused on transportation systems for traffic signaling, public transit, and infrastructure.
The AI City Challenge was one of numerous challenges hosted this week at the Computer Vision and Pattern Recognition (CVPR), which was the second-largest annual AI research conference in 2019 and attracted more than 6,500 participants this year. CVPR also hosted competitions for leading a bot through a RoboTHOR simulated environment, as well as the Deepfake Detection Challenge (although The Register reported the winning team was disqualified).
This year at the AI City Challenge workshop, National Institute of Standards and Technology (NIST) officials detailed plans for the 2020 ASAPS Prize Challenge to create a real-time automated system. That competition will focus on building real-time automated analytics systems for law enforcement and first responders to address real-time emergency event detection like a child falling into a harbor, medical emergencies, abandoned building fires, or other situations.
ASAPS is a multimodal challenge for systems capable of ingesting multiple forms of media, from social media posts and text messages to surveillance camera footage and home video doorbells. ASAPS will feature a series of emergency events in a mock 810-acre city over 24 hours, combining simulated data with physically staged emergencies. The competition will also challenge AI researchers to carry out live video analysis rather than applying the AI to a pre-recorded video.
AI City Challenge organizers said next year it may introduce scenarios involving live video analysis. Dash camera video footage is also under consideration.
The addition of a synthetic data set of 190,000 images for the vehicle reidentification challenge was a unique new part of the AI City Challenge this year. Benchmark data sets used in this year’s competition came from footage provided in part by the Iowa Department of Transportation. Nvidia curated data sets used in the challenge.
In related news, opponents of institutional racism demanding lawmakers defund the police in recent weeks have criticized surveillance systems like the kind championed at the AI City Challenge. Earlier this week, New York City lawmakers passed the POST Act requiring the nation’s largest police department to share what surveillance technology it uses.
If your business is sales, your number one asset is your customers. And customer satisfaction will determine your ultimate success or failure. There are many philosophies and many tools promoting sales efficiency, and in a competitive marketplace, it’s vital to choose the tools that add value and efficiency to your business.
AI tools could improve fake news detection by analyzing users’ interactions and comments
In a paper published on the preprint server Arxiv.org, researchers affiliated with Microsoft and Arizona State University propose an approach to detecting fake news that leverages a technique called weak social supervision. They say that by enabling the training of fake news-detecting AI even in scenarios where labeled examples aren’t available, weak social supervision opens the door to exploring how aspects of user interactions indicate news might be misleading.
According to the Pew Research Center, approximately 68% of U.S. adults got their news from social media in 2018 — which is worrisome considering misinformation about the pandemic continues to go viral, for instance. Companies from Facebook and Twitter to Google are pursuing automated detection solutions, but fake news remains a moving target owing to its topical and stylistic diverseness.
Building on a study published in April, the coauthors of this latest work suggest that weak supervision — where noisy or imprecise sources provide data labeling signals — could improve fake news detection accuracy without requiring fine-tuning. To this end, they built a framework dubbed Tri-relationship for Fake News (TiFN) that models social media users and their connections as an “interaction network” to detect fake news.
Interaction networks describe the relationships among entities like publishers, news pieces, and users; given an interaction network, TiFN’s goal is to embed different types of entities, following from the observation that people tend to interact with like-minded friends. In making its predictions, the framework also accounts for the fact that connected users are more likely to share similar interests in news pieces; that publishers with a high degree of political bias are more likely to publish fake news; and that users with low credibility are more likely to spread fake news.
To test whether TiFN’s weak social supervision could help to detect fake news effectively, the team validated it against a Politifact data set containing 120 true news and 120 verifiably fake pieces shared among 23,865 users. Versus baseline detectors that consider only news content and some social interactions, they report that TiFN achieved between 75% to 87% accuracy even with a limited amount of weak social supervision (within 12 hours after the news was published).
In another experiment involving a separate custom framework called Defend, the researchers sought to use as a weak supervision signal news sentences and user comments explaining why a piece of news is fake. Tested on a second Politifact data set consisting of 145 true news and 270 fake news pieces with 89,999 comments from 68,523 users on Twitter, they say that Defend achieved 90% accuracy.
[W]ith the help of weak social supervision from publisher-bias and user-credibility, the detection performance is better than those without utilizing weak social supervision. We [also] observe that when we eliminate news content component, user comment component, or the co-attention for news contents and user comments, the performances are reduced. [This] indicates capturing the semantic relations between the weak social supervision from user comments and news contents is important,” wrote the researchers. “[W]e can see within a certain range, more weak social supervision leads to a larger performance increase, which shows the benefit of using weak social supervision.”
Let’s block ads! (Why?)
Big Data – VentureBeat