Facebook researcher havedeveloped a new toolthat maps 2D images onto humans in video . The 2D images themselves await featherbrained , but they ’re proof of conception that AI can be trained to face past environmental noise and identify humans in complex scene .
Facebook ’s AI inquiry teamreleased a theme last weekdetailing the process . The squad used a neural net that can analyze complicated scene in picture footage — such as masses walking , dancing , playing basketball — and identify which pixels belong to humans . That ’s step one .
Researchers then overlay all human - belong to pixel with these Minecraft - esque skins . It seems silly , but the grain stay uniform in the videos , even as hoi polloi kept moving and walking . The AI can map the texture onto human race , which is a exploit considering how the people in the video move unpredictably ( imagine what dancing “ looks ” like to a data processor ) , and wear vesture , which alters their soma .
The team discuss gaming lotion in the paper , particularly in the augmented realism space , which Facebook has carry lots of interest in before . But , as The Next Web note , law enforcement would be very concerned in being able-bodied to analyse footage and instantly isolate and categorize people in crowds agree to their movements . dead body cam maker Axon ( formerly bang as Taser ) has been working on similarobject realisation techsince 2017 . For now , at least , the takeaway remain the same : computers are able-bodied to “ see ” us better than ever .
[ The Next Web ]
Daily Newsletter
Get the best tech , science , and culture news in your inbox daily .
News from the future tense , fork up to your present tense .
Please select your desired newssheet and submit your electronic mail to advance your inbox .
You May Also Like
