[NetEase Smart News, August 29] Over the years, we've all become accustomed to seeing footage from security cameras, police surveillance, live streams, and even strangers' social media posts. Even if we wear cameras on our heads, there's no guarantee that someone won't accidentally block or tamper with the device. Imagine searching for a specific individual in months of surveillance footage—such a task would be akin to finding a needle in a haystack. It’s impractical, time-consuming, and incredibly labor-intensive. But for robots, this isn’t an issue.
In the world of Hollywood thrillers, identifying individuals and their actions in videos and images has long been achievable. Tools are now being developed to help us identify people and analyze their activities within visual content. Companies like Facebook and Baidu have been at the forefront of developing this AI technology. As error rates continue to decline and the range of applications grows, we can anticipate that soon every video will be analyzed to extract detailed information about the characters, objects, and actions within.
For decades, artificial intelligence researchers have been working to develop algorithms capable of recognizing images and interpreting their contents. The complexity of images—each composed of millions of pixels—creates unique patterns far too intricate for traditional hand-coded algorithms to handle effectively.
In 2012, researchers introduced a technique called deep learning, which mimics the interconnected neurons in our brains. This approach works exceptionally well when dealing with vast quantities of images. By training a deep neural network with sufficient examples, the system can detect shared patterns across different images, such as the distinct features of various breeds of cats.
Since then, these systems have grown significantly in scale and complexity. Researchers began building larger "neural networks," while hardware manufacturers like NVIDIA started designing specialized processors to accelerate the training process. As a result, the capabilities of these systems have skyrocketed. With large amounts of image or video data, these systems can be trained to understand appearances and consistently identify individuals with remarkable accuracy.
A notable example is the MegaFace dataset at the University of Washington. This dataset contains nearly five million images of 672,000 people sourced from Flickr. In July, the MegaFace team showcased the latest performance metrics of algorithms trained on this dataset. When matching two faces of the same person from two separate datasets of one million photos, the top-performing teams achieved an accuracy rate of 75% with one attempt and over 90% with ten attempts.
"We need to conduct global testing of facial recognition capabilities to make practical applications a reality," said Ira Kemelmacher-Shlizerman, the University of Washington professor overseeing MegaFace, during an interview with the university press. "Large-scale testing helps us identify both the flaws and successes of these recognition algorithms."
Video analysis employs similar technology, capturing still images while requiring higher computational power. This enables AI systems to understand events unfolding over time. At the end of August 2017, Baidu announced winning the "ActivityNet" challenge, successfully tagging human behaviors in 300,000 videos with an accuracy rate of 87.6%. Examples include tasks like chopping wood, cleaning windows, and walking dogs.
Facebook has also shown interest in this technology, aiming to understand who appears in real-time videos and what they're doing. In a recent interview, Joaquin Piñero-Hotandela, Director of Applied Machine Learning, mentioned that ideally, Facebook would analyze every live stream to offer personalized video feeds to users.
The U.S. government has already begun implementing this technology in limited capacities. Last week, the New York Department of Motor Vehicles reported arresting over 4,000 individuals using facial recognition software. Unlike scanning surveillance footage, the software compares driver’s license application photos with existing images in databases, making it harder for fraudsters to impersonate others. Expanding this technology publicly could potentially cover over 50% of American adults, with larger datasets leading to improved AI performance.
This trend may not be far off. Axon, known for its Tasers, is the largest supplier of cameras to U.S. police agencies. Recently, the company has expressed ambitions to integrate AI into its products. Earlier this year, Axon acquired two AI firms. CEO Rick Smith previously told Quartz that the ideal use case for AI would involve generating incident reports objectively, freeing officers from paperwork. While facial recognition remains largely passive today, it holds immense potential for the future.
Motorola, another key provider of cameras, highlights its software's ability to quickly learn facial expressions, aiding police searches for missing children.
Security cameras are also driving advancements in AI technology. In April, Intel announced hardware for security cameras capable of "dense monitoring, stereoscopic vision, face recognition, counting," and "behavioral analysis." The product claims to be waterproof, self-sufficient, and "almost indestructible," allowing operation in remote areas without internet access, including identifying returning customers at cash registers.
But what about privacy-conscious citizens concerned about surveillance? There are ways to fool facial recognition software, such as wearing specially designed glasses or applying graffiti to mislead AI. However, this often requires understanding how these algorithms function. (From: Quartz Media; Editors: Wu Man)
As AI technology evolves, it's crucial to balance innovation with ethical considerations, ensuring that privacy rights remain protected in an increasingly connected world.
Logic Gates And Inverters
Logic Gates And Inverters,Circuit Logic Gates And Inverters,Logic Chip Gates And Inverters,Gates And Inverters Logic Output
Shenzhen Kaixuanye Technology Co., Ltd. , https://www.icoilne.com